Moral Communities and the Deterrence ... | Wiley Online Library

Religion and Gambling Among Young Adults in the United States: Moral Communities and the Deterrence Hypothesis DAVID EITLE Department of Sociology & Anthropology Montana State University

Despite voluminous research examining religion as an integrative force and a mechanism of social control, relatively few studies have examined the association between religion and proscribed or morally ambiguous behaviors beyond crime and drug use. The present exploratory study examines the role of religion, at both the individual and county levels, in predicting self-reported gambling problems. Hierarchical linear models are employed to examine religion and self-reported gambling problems using the restricted use data of the National Longitudinal Study of Adolescent Health. A negative association between religious attendance and problem gambling (at the individual level) is strongest when church adherents per capita is relatively high (measured at the county level). However, when the number of conservative Protestants per capita is relatively high, religious attendance (measured at the individual level) is associated with an increased risk of gambling problems. These countervailing findings are interpreted as supportive of the bonding and bridging capital thesis.

Keywords: gambling, moral communities, deterrence hypothesis, young adults.

INTRODUCTION Sociologists have long explored the salience of religion as both an integrative force and a social control mechanism. Extant research has demonstrated support for the notion that religion serves as a protective factor against participation in illicit and immoral behavior (Bahr et al. 1998; Baier and Wright 2001; Burkett 1993; Johnson et al. 2000; Stark and Bainbridge 1997). However, many of these inquiries have been limited in a few important ways. First, while many studies have examined the effects of an individual’s religious commitment and/or participation on criminal behavior, there are relatively few studies that have explored the influence of religion on other forms of proscribed or morally ambiguous behavior beyond crime. Indeed, one historically denounced act, gambling, and its potential association with religion has been described as an understudied relationship (Hoffmann 2000).1 This is particularly surprising given both the welldocumented adverse consequences of some gambling behaviors, such as problem gambling (e.g., Grinols 2004; Gupta and Derevensky 1998; Korn and Shaffer 1999), and given the belief that behaviors such as gambling, as a “nonvictim” act, may be the type of activity that the influence of religion is most likely to deter (Burkett and White 1974). Second, relatively few studies have examined the association between religion and either criminal or proscribed behavior at the contextual level, despite the existence of compelling theoretical claims that religious institutions serve to create a moral community that reduces community-level crime and immoral activity by both increasing integration and increasing social control (Cochran and Akers 1989; Kent and Doyle 1982; Lee and Bartkowski 2004; Stark 1996; Stark, Doyle, and Kent 1980). And while a few studies have examined the association between Correspondence should be addressed to David Eitle, Department of Sociology & Anthropology, Montana State University, Wilson 2-126, PO Box 172380, Bozeman, MT 59717-2380. E-mail: [email protected] 1

See Ellison and McFarland (2011), this issue, for another recent treatment of religion and gambling.

Journal for the Scientific Study of Religion (2011) 50(1):61–81 2011 The Society for the Scientific Study of Religion

C

62

JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION

contextual-level religion and aggregate indicators of criminal behavior (i.e., crime rates; e.g., see Bainbridge 1989; Beyerlein and Hipp 2005; Desmond, Kikuchi, and Morgan 2010; Lee 2006; Lee and Bartkowski 2004), examinations of the influence of the larger, contextual-level measures of religion on morally disapproved behaviors such as gambling has largely been neglected.2 Finally, only a few published studies have examined the role of both the contextual and individual indicators of religion in predicting illegal and proscribed behaviors (e.g., criminal behavior— Regnerus 2003; adolescent alcohol use—Bjarnason et al. 2005). No published study has explored both the contextual and individual effects of religion and gambling behavior. The present study aims to extend our current understanding of the association between religion and gambling in three important ways. First, this study represents one of the few efforts to examine the contextual-level influence of religion on gambling behavior, examining the salience of various theoretical notions regarding the role of the larger context and religion on anti-ascetic behaviors such as gambling. Second, it examines the potential interactions between contextuallevel measures of religion and individual religiosity to evaluate whether such contextual factors moderate the relationship between individual religiosity and gambling behavior. And finally, this study will also examine the role of other contextual and individual factors that have been proposed to increase the likelihood of gambling—few studies have comprehensively examined both sets of factors simultaneously when examining gambling behavior. BACKGROUND Gambling in America has increased dramatically over the past two decades, largely ushered in with the tremendous growth in legalized gambling. Indeed, Grinols (2004) documents that at the start of the 20th century, 35 states had constitutional bans in effect on some form of gambling. But today, several forms of gambling are now sanctioned (or even conducted) by the state—42 states have sponsored lotteries (Clotfelter and Cook 2009), 32 states have legal casino gambling or card rooms (Welte et al. 2002), and pari-mutuel racing of some sort is available in 45 states (Thompson, Gazel, and Rickman 1996). Ninety-five percent of Americans now live in a state with a lottery (Melnick 2009) and 48 states have some form of legal gambling (Dombrink 2009). And Internet gambling is a huge endeavor as well as an estimated $6 billion industry in 2005 (Wolfe and Owens 2009:5). According to recent research, the majority of adults in America admit to gambling in the past year (82 percent according to Welte et al. 2002) and the percentages are substantial among adolescents and young adults as well; Welte et al. (2008) report that 68 percent of respondents in a sample of 14–21 year olds report gambling in the past year, with 11 percent revealing that they have gambled more often than twice each week. And gambling casts a huge economic presence in our society, with legal gambling in America estimated to accumulate revenues of approximately $91 billion annually (American Gaming Association 2007); estimates of the extent of illegal gambling range from one-third to roughly the same as those dollars bet in legal enterprises (Issacs 2001). While these facts obfuscate efforts to define gambling as morally inappropriate behavior, various religions have defined gambling as proscribed activity throughout history. Hoffmann (2000) noted that several faiths that have been characterized as conservative Protestant denominations have fought to curb legalized gambling in the United States. Furthermore, Hoffmann noted that several denominations have forwarded both scriptural-based arguments proscribing gambling as well as producing position papers that warn of the risks of gambling, by summarizing the social science evidence that demonstrates the social and economic costs of gambling 2

The terms “morally disapproved” and “proscribed” are used to describe how various social groups depict gambling; these terms are not used to convey any particular moral judgment about gambling beyond the recognition that such behavior has been condemned by various groups, including various religions, to varying degrees.

63

RELIGION AND GAMBLING

(2000:490). Accordingly, Aasved reviewed this tradition by stating “gambling, whether pathological or non-pathological, has . . . been condemned as an unchristian and uncapitalistic tool of the devil” (Aasved 2003:7). In addition, Burkett and White (1974) argued that behaviors such as gambling, premarital sex, alcohol abuse, and drug use would be the types of acts that are most likely to be deterred by religion—they argued that only when secular values are vague should one expect to see ascetic-based standards to have an effect (when secular rules proscribe a behavior, anti-ascetic values merely reinforce such proscriptions). This notion—called the anti-asceticism hypothesis—has been routinely supported, with several studies finding an inverse association between anti-ascetic acts (especially alcohol use) and religiosity (for a review, see Cochran and Akers 1989:201; see also Benda 1995). Although there have been only a limited number of studies examining the association between religion and gambling behavior, the general thrust of these findings is consistent with the larger body of evidence that shows that individual religiosity serves to reduce various forms of proscribed activity. For example, Hoffmann (2000) found that one measure of religiosity, attendance, was negatively associated with problem gambling in his analysis of cross-sectional data drawn from a nationally representative sample of adults. However, Hoffmann also found that the importance of faith in God was not a significant predictor of problem gambling. Similarly, Diaz (2000) also found that church attendance was negatively associated with both the frequency of gambling and the amount of money gambled among a sample of Las Vegas residents, while Ellison and Nybroten (1999) found that both being a conservative Protestant and church attendance were significant predictors of opposition to state supported lotteries. Finally, Welte et al. (2008) found that although Baptists were more likely to engage in frequent gambling than other religious adherents, there were no significant differences in the associations between religious affiliation and problem gambling. Overall, these studies are illuminating but do not consider the role of the larger religious context (in addition to individual-level factors) as either having a direct effect on an individual’s gambling or as moderators.

THEORETICAL FOUNDATIONS Individual-Level Explanations While a number of generalist explanations of crime/deviance have been applied to understand the relationship at the individual level between religion and gambling, Pearce and Haynie (2004) posit that three explanations are particularly salient. Social learning explanations, from Sutherland’s (1947) differential association theory to Aker’s (1997) reformulation and extension of differential association, emphasize the role of the content espoused in socialization experiences— people develop values systems and behavioral repertoires in response to being taught (by intimate others) the benefits and costs of engaging in various illicit behaviors. According to this theory, people exposed to religious teachings and religious others (from parents to clergy) would be less likely to engage in gambling behavior to the extent that it is condemned by the teachings and by others (Pearce and Haynie 2004:1556; Smith 2003). Hence, the greater the exposure to such religious teachings and religious others, the more likely one is to accept a belief that gambling is morally inappropriate behavior and, thus, the less likely one would be to engage in gambling. Strain-based explanations, such as Agnew’s (1992) general strain theory, suggest that crime and other illicit behaviors are best seen as a coping strategy employed to deal with the strain experienced by individuals. Religion and religious agents may reduce the chances of an individual engaging in gambling by providing him or her with alternative coping strategies to employ in response to strain (Pearce and Haynie 2004; Smith 2003). Similarly, Ellison and Levin (1998) suggest that religion may provide people with greater social capital—resources that may either reduce the amount of strain that an individual experiences or provide alternative, prescribed

64


outlets to utilize to cope with strain. Thus, people who have a greater commitment to religion and who interact with religious others are predicted to be less likely to engage in gambling, due to religion providing an alternative coping strategy to deal with life’s stresses and/or as a means to reduce one’s exposure to life’s stresses. Finally, social control explanations (such as Hirschi 1969) suggest that the extent and strength of social ties between the individual and conventional society serve to reduce both the opportunities to engage in deviant behavior as well as increasing social attachment and adherence to moral codes (Pearce and Haynie 2004:1556). Religious institutions, similar to other important institutions such as family and schools, should serve to deter proscribed behavior generally (Baier and Wright 2001). Social integration is seen as fundamental to increasing the social control function of religious groups, which should serve to reduce the likelihood of gambling, including frequent or problem gambling (Ellison 1991; Hoffmann 2000:491; Umberson 1987). Likewise, the individual who is socially integrated into religious groups may have greater clarity (as well as reinforcement) regarding what is unacceptable behavior than the less integrated. According to Rohrbaugh and Jessor (1975), religion serves as social control mechanism in four ways: (1) by embedding the person in organized sanctioning networks; (2) by increasing sensitivity to what constitutes moral and appropriate behaviors; (3) by offering the deity as the supreme source of wrath and punishment; and (4) by creating an obedience orientation through encouraging devoutness (1975:137; as cited in Rostosky et al. 2004:682). The last two ways—reducing deviant behavior by the threat of deity punishment and the promise of deity reward—have commonly been referred to as the hellfire hypothesis of Hirschi and Stark (1969). Despite the different mechanisms that are articulated by these explanations, each posits the same outcome—greater religiosity will serve to reduce prohibited behavior. And recent metaanalyses of the extant research reported that the majority of studies find that religiosity serves to deter crime and other proscribed activity (Baier and Wright 2001; Johnson et al. 2000), a prediction generally referred to as the deterrence hypothesis. These ideas spawn the following hypothesis for the religion-gambling problems relationship: Hypothesis 1: Individuals who are more religious will report fewer (or no) gambling problems. Beyond these generalist explanations, one additional hypothesis that is salient for assessing the association between an individual’s religion and his/her gambling behavior is the norm qualities hypothesis (Cochran and Akers 1989). According to the norm qualities hypothesis, some religious denominations have greater proscriptive qualities than others, especially regarding behaviors that may not be defined normatively as proscribed activity. And of the limited research that has examined the association between religious heritage and illicit behavior, the findings are largely supportive of the general notion that conservative Protestants are less likely than others to report engaging in such behavior as alcohol use and drug use (Cochran and Akers 1989; Jensen and Erickson 1979; Nelson and Rooney 1982). In addition, religious denominations may vary in their condemnation of gambling. Catholicism, for instance, does not discourage gambling to the extent that other Protestant faiths do (Abbott and Volberg 2000; Edmondson 1986) and endorses charitable gambling, notably bingo. Furthermore, Welte et al. (2008) found that respondents in the category “other than Baptist” Protestants were more likely (along with Catholics) to participate in gaming activities in the past year. And some research has found evidence of a significant interaction between individual measures of religion and religious denomination, such that the influence of such measures as religiosity and religious attendance on deviant behavior is strongest among conservative Protestants (Nelson and Rooney 1982; see also Jensen and Erickson 1979). Hypothesis 2: Conservative Protestants will report fewer (or no) gambling problems, relative to others.


65

Hypothesis 3: The effect of religiosity and religious attendance will be stronger for conservative Protestants than those affiliated with other denominations. Contextual-Level Explanations At the macrolevel, the earliest explanation of how religion reduces proscribed activity is referred to as the moral communities hypothesis (Stark, Doyle, and Kent 1980; Stark, Kent, and Doyle 1982). As it was originally conceptualized, this thesis suggested that the strength of the religious institutions in the larger context engendered a sort of moral ecology that would condition the individual-level association between religiosity and illicit behavior—in communities wherein the majority of constituents are involved in religion, the individual-level effect of religiosity on proscribed behavior should be greatest (Stark and Bainbridge 1997). Regnerus (2006) refers to this conditional association between individual and contextual religion as “the light switch portion of the moral communities thesis”: only when a religious individual is in community with a critical mass of others who share their beliefs and practices will that individuals’ religious beliefs significantly affect their behavior. As Stark (1996:164) puts it—“what counts is not only whether a particular person is religious, but whether this religiousness is, or is not, ratified by the social environment.” Communally ratified religiosity, in essence, “turns on” the light switch of an individual’s own personal belief system. (2006:268)

Conversely, those communities that have relatively fewer members involved in religion are expected to have a weaker association between individual religion and deviant behavior, such that its association will be (relatively) modest, or even nonexistent (Regnerus 2003). More recently, this hypothesis has also been interpreted to posit a direct association between the strength of the pro-religious climate at the larger contextual level and illicit behavior rates (Lee 2006; Regnerus 2003; Stark 1996). Religious communities are seen to promote conformity in a number of ways, including by increasing social organization, promoting collective efficacy (shared expectations of informal control of the children of the community) among communities and by reducing behavioral options (Lee 2006:313). And while some studies have found either weak or no support for the moral communities hypothesis (Cochran and Akers 1989), more recent research has found a direct negative association between contextual-level indicators of religiosity and illicit behavior, including such activities as violent crime (Lee 2006). Of the two recent studies that have evaluated the direct effect of the larger religious context as well as the potential moderating effect that a moral community has on the individual-level religiosity-illicit behavior relationship, both studies found support for the notion that a stronger moral community strengthens the association between an individual’s level of religiosity and illicit activities, including crime (Regnerus 2003) and adolescent alcohol use (Bjarnason et al. 2005). Hypothesis 4: Respondents residing in contexts with a greater concentration of religious adherents will report fewer (or no) gambling problems. Hypothesis 5: The strength of the association between individual-level religiosity and reported gambling problems should be magnified under conditions of a greater concentration of religious adherents at the contextual level. While the moral communities thesis, as originally formulated (Stark and Bainbridge 1997), does not espouse that the nature of the specific denominational affiliation matters, another contextual-level explanation, the conservative Protestantism thesis, suggests that the affiliation of religious adherents in the larger context does matter. This thesis suggests a notion that is akin to the norm qualities thesis forwarded to explain the individual-level association between

66


religion and illicit behavior. This thesis has been proffered by some (e.g., Regnerus 2003) to explain how the extent of conservative Protestantism at the macrolevel can reduce criminal and immoral behavior because of the greater proscriptiveness of the value system of conservative Protestants, with Regnerus (2003) finding that at the contextual-level conservative Protestantism was associated with lower theft and minor delinquency counts. If the thesis has merit for the present study, the following hypothesis should be supported: Hypothesis 6: Respondents residing in contexts with a greater concentration of conservative Protestant religious adherents will report fewer (or no) gambling problems. Hypothesis 7: The strength of the negative association between individual-level religiosity and reported gambling problems should be strengthened under conditions of a greater concentration of conservative Protestant religious adherents at the contextual level. A third contextual-level explanation, the bonding/bridging capital thesis of Beyerlein and Hipp (2005), may also be relevant to understanding the association between religion and gambling. These authors argue that the relative size of the conservative (evangelical) Protestant population at the macrolevel will be associated with higher rates of crime, not because of the promotion of certain values, but rather because these religious groups are more likely to promote the formation of bonding capital (i.e., generating/reinforcing social ties and networks among primarily the religious group members themselves), which they suggest will be less effective in generating social ties and networks that would help the larger aggregate reduce crime.3 However, these authors suggest (and their research supports their thesis) that other types of religious heritages, namely, mainline Protestants and Catholics, promote the formation of another type of capital, bridging capital, that effectively extends linkages between the religious community and the larger aggregate and serves to increase social capital throughout the area, and that facilitates efforts to reduce crime. If one extends this thesis to gambling behaviors, it can be expected that gambling would be more of a problem in areas with relatively larger conservative Protestant populations (contrary to the conservative Protestantism thesis) but may be less of a problem in areas with relatively large religion-adhering populations overall. Hence, this idea both supports the aforementioned hypotheses (no. 4 and no. 5) derived from the moral communities notion and the following hypothesis: Hypothesis 8: The strength of the negative association between individual-level religiosity and reported gambling problems should be weakened under conditions of a greater concentration of conservative Protestant religious adherents at the contextual level. In summary, while there are compelling reasons to believe that individual religiosity and macro-level religious participation should be associated with gambling behavior, there has been little extant research that has explored these possible connections. The present study represents an exploratory study into these connections in order to begin to understand how such larger contexts may both directly affect individual gambling and condition the relationship between religiosity and gambling at the individual level.

3

Other scholarship (e.g., Ellison, Burr, and McCall 2003; Lee 2006) has also forwarded the notion that the extent of conservative Protestantism can serve to increase the extent of violent criminal behavior at the macrolevel (because of the emphasis on values that tolerate violence used in defense of honor, property, or family and women). However, such a thesis (promotion of values that tolerate violence) does not easily extend to gambling.

67


METHODS Data The data for this analysis come from the restricted use sample of the National Longitudinal Study of Adolescent Health (Add Health). The Add Health data, a nationally representative sample of American adolescents (initially in grades 7–12 when the first wave of data was collected) were gathered by employing a school-based, clustered sampling design (Bearman, Jones, and Udry 1997). The data provide detailed information on a variety of behaviors, including religious beliefs and behaviors, that have spawned recent studies examining the myriad ways religion can influence behavior (for recent examples, see Burdette et al. 2009; Regnerus 2003). The present analysis uses data collected in wave 3 when the respondents were adults, between the ages of 18–27, because this was the only wave in which questions on gambling activity were asked. However, examining the link between gambling and religion among young adults is also a strength of the present study, since most studies examining the association between religion and deviance have targeted adolescents (Hoffmann and Bahr 2006). Consistent with past research testing the effects of the larger religious context on various forms of deviant/criminal behavior (Bellair, Roscigno, and McNulty 2003; Beyerlein and Hipp 2005; Lee 2006; Lee and Bartkowski 2004; Regnerus 2003), the present study employs counties as the unit of analysis.4 County-level data from the 2000 U.S. Census and data collected by the Association of Statisticians of American Religious Bodies and published by the Glenmary Research Center (Bradley et al. 1992) are employed to capture both the county-level religion variables and other salient structural variables. Dependent Variable The dependent variable, gambling problems, is a sum of four dichotomous measures of selfreported gambling problems. Questions utilized to measure gambling problems included asking about whether one: (1) spent a lot of time (periods two weeks or longer) thinking about gambling experiences or planning out future gambling ventures/bets; (2) ever gambled to relieve uncomfortable feelings such as guilt, anxiety, helplessness, or depression; (3) had ever lost money gambling one day and returned another day to get even; and (4) experienced serious relationship problems due to gambling. Higher scores in this scale that ranged from 0 to 4 indicate greater gambling problems (Cronbach’s alpha = .760).5 Since approximately 28 percent of respondents in the sample did not report ever gambling (a second measure asked about lifetime gambling), the analyses predicting gambling problems were reanalyzed with the inclusion of only those reporting that they had ever gambled; the results were not fundamentally different than the reported analyses that include those individuals (coded as zero for the gambling problems; results available upon request). The measure of gambling problems that is utilized in the present study should not be confused with a clinical defined measure of gambling problems. The use of a well-established measure of problem gambling, such as the South Oaks Gambling Screen (Lesieur and Blume 1987), the NORC DSM Screen for Gambling Problems (NODS)6 (Gerstein et al. 1999), or 4

There are a couple of reasons for using counties as the aggregate unit of analysis. Land, McCall, and Cohen (1990) have argued that the level of aggregation in deviance research is arbitrary because of the convergence in studies’ findings across various units of analysis. Furthermore, the data available to measure the key indicators of religion at the aggregate level are only found at the county level (Lee 2006). Despite these reasons, it should be noted that other aggregations smaller than counties may produce different findings than those reported.

5

While it is important to acknowledge the definitional problems associated with defining problem or pathological gambling, the present study does not claim to provide a clinical categorization of problem gamblers.

6

NORC is the National Opinion Research Center; DSM stands for Diagnostic and Statistical Manual.

68


the DSM of Mental Disorders scale (American Psychiatric Association 1994; see also Svetieva and Walker 2008) with established psychometric properties may produce different results than those uncovered in these analyses. However, the percentage of respondents in the current study who report having experienced at least one gambling problem (2 percent) is in range with other studies that have employed more established measures. For example, using the NODS, researchers at the NORC (Gerstein et al. 1999) found a lifetime rate of problem gambling of 1.5 percent for adults (an additional .9 percent were classified as pathological gamblers); using pathological gambling symptoms items from the DSM-III and DSM-IV, Slutske, Jackson, and Sher (2003) found lifetime prevalences of problem gambling (defined as reporting at least one symptom of pathological gambling) of between 3.2 percent and 5.3 percent across three waves of data collection among a sample of 18–29 year olds (similar to the age range of the present study’s sample). Nonetheless, the analyses should be treated as exploratory in nature due to this rudimentary measure of gambling problems. Individual-Level Variables The analysis included two key individual-level variables: religiosity and religious attendance. A measure of religiosity is constructed from scores derived from a factor analysis with varimax rotation of five items capturing both attitude and practice outside of church: (1) how important is your religious faith to you (four-item response ranging from “not important” to “more important than anything else”); (2) I am being “led” spiritually (five-item response ranging from “strongly agree” to “strongly disagree”); (3) how often do you pray (outside of church, synagogue, etc.; eight-item ordinal response ranging from “never” to “more than once a day”); (4) to what extent are you a religious person (four-item response ranging from “not religious at all” to “very religious”); and (5) I employ my religious or spiritual beliefs as a basis for how to act and live on a daily basis (five category response ranging from “strongly agree” to “strongly disagree”) (Cronbach’s alpha = .789).7 Higher scores indicate greater religiosity. Religious attendance was measured by asking the respondent how often in the past 12 months he or she attended religious services. And an indicator of the religious denomination that the respondent adheres to is also considered. The six categories (Catholics, moderate Protestants, conservative Protestants, other Protestants, other religious affiliation, and the nonaffiliated) roughly approximate Steensland et al. (2000) recommendations. However, due to concerns with collinearity, I do not break out black Protestants because of its strong association with race (see Adamcyzk and Felson 2008; Hill 2009).8 In addition to the measures capturing religion, a variety of controls are also considered. An index of deviant behavior is constructed from 12 items, ranging from passing bad checks to being stopped by the police.9 Prior research has found that deviant behavior is a valid predictor of

7

An exploratory factor analysis of the religion items included in the Add Health data led to the selection of these five items as an indicator of religiosity. An additional factor analysis of each of the aforementioned items and the religious attendance measure indicated that while the five religiosity items loaded strongly on one factor, the religious attendance measure loaded more strongly on a second factor (eigenvalues > 1), indicating it is capturing a different dimension. In addition, there exists some evidence that religious attendance (and other religious activities) may have greater predictive utility with deviant behavior than other measures of religion (Evans et al. 1995). Hoffmann and Bahr (2006:244) speculate that participation in religious activities may be a more effective deterrent due to the social networks and moral messages instilled during such activities.

8

Because of the small number of people identifying themselves as Jewish, these respondents were included in the residual category “other religious affiliation.” The specific denominational affiliations that comprise the six categories are available upon request from the author. 9 Other deviant acts included in this measure are grand theft, burglary, robbery, selling drugs, petty theft, assault, buying stolen goods, credit card fraud, and aggravated assault. Cronbach’s alpha = .70.


69

gambling activity among youth (Gupta and Derevensky 1998; Langhinrichsen-Rohling et al. 2004). Other demographic variables are also considered, including race/ethnicity (African American, non-Hispanic white, and other racial groups), gender (1 = male), age, and whether the respondent is currently employed or not. Two measures of education are also considered: (1) a measure capturing how many years of education the respondent had completed by the time of the interview; and (2) a dichotomous measure capturing whether the respondent is currently enrolled in college or other school. Melnick (2009:105) notes that high school dropouts spend on average four times as much on lottery tickets as college graduates. Contextual-Level Variables Three key county-level measures of religion are used: (1) church adherents per capita (total); (2) conservative Protestant adherents per capita; and (3) a measure of religious heterogeneity. While the first two measures are used to capture the moral communities and conservative Protestantism hypotheses, respectively, the third measure—religious heterogeneity—is used as a measure to evaluate whether the diversity in denominations in a county serves to weaken the overall strength of the larger religious context.10 McVeigh (2006:519) suggests that while religious heterogeneity provides opportunities for interfaith relationships, it likely undermines the social integration and moral community among residents (see also Ellison, Burr, and McCall 1997). Indeed, McVeigh (2006) and Ellison et al. (1997) found that religious heterogeneity was associated with crime and suicide rates. In addition, past research that has explored the macro-level predictors of deviant behavior generally, and gambling behavior specifically (Welte et al. 2004), has examined other structural measures that are included in these analyses. Two common indicators of the degree of economic disadvantage experienced by a county are employed: the county poverty and unemployment rates. Since prior research has demonstrated that such structural factors tend to be highly correlated (Land, McCall, and Cohen 1990), a principal components analysis was employed to reduce problems of multicollinearity and the partialling fallacy and as a data reduction strategy. The results of such an analysis led to the construction of an index of disadvantage that summed the standardized measures of the two variables. Additional structural controls that may be associated with both the religion measures and gambling activity that are included are the proportion of black residents, the proportion of Hispanic residents, and the proportion of college graduates. Analysis Strategy A hierarchical linear modeling procedure is employed to generate parameter estimates because the data are multilevel (Bryk, Raudenbush, and Congdon 1996). There are a number of methodological advantages for using multilevel modeling. First, multilevel models do not violate the assumption of independence among observations because they explicitly recognize the clustering of individuals within higher-level units such as counties. Second, hierarchical models are advantageous for estimating cross-level effects, since all estimates are adjusted for the covariates, despite whether they are measured at the individual or contextual level. And third, hierarchical models both partition the variance between levels and also statistically separate the variance of the individual-level parameters from sampling variance. The inability to factor out the sampling variance when data are hierarchical results in an underestimation of the explanatory power of contextual variables. All level 2 (county) measures and the level 1 variables (except for The religious heterogeneity measure is calculated as 1 − Pi2 , where Pi is the proportion of the total adherents in each of the following religious categories: Jewish, Catholic, Black Baptist, “other” conservative, “other” moderate, and “other” liberal).

10

70


dichotomous measures) are grand mean centered prior to analyses (Kreft, De Leeuw, and Aiken 1995). Because of the unequal probability of sample selection (Chantala 2006), all analyses were weighted to account for the Add Health design effects. The sample analyzed consists of data from wave 3 of the in-home interviews conducted with a sample of the respondents who had valid weights and who answered in-depth questions on a variety of sensitive issues, including gambling behavior (only measured in wave 3). Finally, only respondents who resided in a county with at least four other respondents (in the sample) were included in these analyses.11 Modeling equations with counties that have few respondents nested in them becomes a daunting task, and it is a common strategy in multilevel methods to exclude cases in which the number of level 1 units (in this case, respondents) is less than five (e.g., Lee et al. 1998). Employing such strategies produced a sample of 9,320 respondents across 102 counties. The basic analysis strategy employs Poisson regression models. Poisson models are commonly employed to analyze count data.12 I examine four models: (1) a model considering individual-level measures only; (2) a model considering both individual- and county-level measures; (3) a model including interaction terms that examine the potential moderating role that county-level measures of the moral community influence the strength of the associations between religious attendance, religiosity, and the dependent variables; and (4) a model that includes an interaction term between the county-level measure of conservative Protestantism and the respondent’s conservative Protestant affiliation. The intercept, β oj , represents the mean level of reported gambling problems in county j when all other level 1 variables are at their means and when the dichotomous variables values equal zero. Table 1 provides the descriptive statistics for each of the variables considered in the analyses.

RESULTS In Table 2, I present the results of models predicting the number of gambling problems reported by the respondents. In model 1, I only include individual-level measures, and the results indicate that a number of demographic factors are associated with reported gambling problems. Males and older respondents are more likely to report gambling problems (and more of them), while Hispanics (relative to non-Hispanic whites) are less likely to report such problems. Being male increases the expected number of gambling problems by a factor of 4.31, holding all other variables constant. One other control, self-reported deviant behavior, was also found to be a statistically significant predictor of the number of reported gambling problems. Respondents who participate in more deviant activities are also predicted to report more gambling problems, holding all other variables constant. Specifically, a unit increase in reported deviant activity

11

Because the sampling design strategy of the National Longitudinal Study of Adolescent Health targeted schools as the primary sampling unit, individual respondents who resided in counties in wave 3 that had less than four other respondents had likely moved from the county where he/she attended school in wave 1 of the data collection (AddHealth tracked and attempted to collect information on respondents who moved). Sensitivity analyses considering these excluded respondents suggest that there are no significant differences between the groups on such key characteristics as gambling problems, the religiosity index, gender, hours worked, or self-reported deviance. Additional analyses (available upon request) examining individual-level factors that included these respondents revealed little substantive differences in the pattern of significant findings across this model and the analogous model reported in Table 2.

12

Parameter estimation in HLM with models predicting count data can be accomplished by using either penalized quasilikelihood (PQL) or Laplace 6 estimation approaches. While PQL is widely implemented (it is the default estimation in HLM), PQL yields biased estimates with Poisson data when the mean number of counts is less than five (Bolker et al. 2008). Furthermore, PQL estimation computes a quasi-likelihood rather than a true likelihood, which renders comparisons across nested models inappropriate. In addition, Laplace approximations have been found to produce accurate estimators for Poisson mixed models with overdispersed count data (Joe 2008).

71


Table 1: Descriptive statistics (n = 9,320 respondents; 102 counties) Variable Individual Level Gambling problems Religious attendance Religiosity Catholic No religious affiliation Moderate Protestant Conservative Protestant Other Protestant Other religious affiliation Gender (male = 1) Age Deviant behavior Non-Hispanic white Hispanic/Latino African American Other racial group Education In school (1 = yes) Employed (1 = yes) County Level Church adherents (per capita) Religious heterogeneity Conservative Protestant adherents (per capita) Proportion black Proportion Hispanic Proportion college graduates Structural disadvantage Midwest South West Northeast

Mean/Proportion

Std. Dev.

Min

Max

.03 2.12 .00 .26 .20 .11 .29 .05 .05 .47 21.66 1.07 .57 .16 .20 .08 13.24 .41 .74

.27 1.95 .95 .44 .40 .32 .46 .22 .21 .50 1.61 2.23 .50 .36 .40 .26 1.90 .49 .44

0 0 −2.11 0 0 0 0 0 0 0 18 0 0 0 0 0 6 0 0

4 6 1.82 1 1 1 1 1 1 1 27 35 1 1 1 1 22 1 1

.55 .73 .19

.15 .14 .16

.20 .36 .02

.97 .96 .66

.13 .06 .24 −.11 .25 .40 .19 .17

.15 .10 .08 1.78 .43 .49 .39 .37

0 .001 .08 −2.47 0 0 0 0

.59 .65 .45 4.84 1 1 1 1

is associated with approximately a 12.75 percent increase in the expected number of reported gambling problems, holding other variables constant. Of the religion variables, neither the measure of religiosity nor the religious attendance measure is significantly associated with reported gambling problems. Hence, I find little support for the religious deterrence hypothesis (hypothesis no. 1) with regard to gambling problems among this sample of young adults. However, one’s particular religious affiliation does appear to have an association with reported gambling problems: relative to the referent category (the nonaffiliated), Catholics, and those in the category of “other religious affiliation” appear to be at an increased risk of reporting gambling problems. Being Catholic (as opposed to having no religious affiliation) increases the expected number of gambling problems by a factor of 4.31, holding all other variables constant. And since conservative Protestant affiliation was not found to have a statistically significant association with gambling problems, one cannot evince support from this analysis for the moral qualities hypothesis (hypothesis no. 2).

72


Table 2: Multilevel Poisson regression of gambling problems (n = 9,320 respondents; 102 counties) Variable Individual Level Religious attendance Religiosity Catholic Moderate Protestant Conservative Protestant Other Protestant Other religious affiliation Gender (male = 1) Age Deviant behavior Hispanic/Latino African American Other racial group Education In school (1 = yes) Employed (1 = yes) Religious attendance × Catholic County Level Intercept Church adherents (per capita) Religious heterogeneity Conservative Protestant adherents (per capita)

1

2

3

4

−.08 (.10) −.05 (.16) .97∗∗ (.35) .81 (.54) .12 (.43) 1.02 (.56) 1.00∗ (.40) 1.46∗∗ (.29) .24∗∗ (.07) .12∗∗ (.01) −.97∗∗ (.34) .25 (.28) .59 (.39) −.03 (.10) −.55 (.34) −.25 (.29)

−.05 (.13) −.05 (.16) 1.52∗∗ (.39) .66 (.54) −.08 (.45) .80 (.57) .85 (.47) 1.47∗∗ (.29) .24∗∗ (.07) .12∗∗ (.01) −.99∗∗ (.34) .22 (.28) .64 (.39) −.02 (.10) −.56 (.34) −.25 (.29) −.44∗∗ (.13)

.05 (.13) −.05 (.17) 1.57∗∗ (.43) .68 (.56) −.07 (.47) .76 (.53) .90 (.48) 1.46∗∗ (.30) .25∗∗ (.08) .12∗∗ (.02) −1.10∗∗ (.38) .24 (.30) .55 (.39) −.02 (.10) −.58 (.34) −.25 (.30) −.45∗∗ (.14)

.26 (.20) −.54 (.50) 1.49∗∗ (.37) .67 (.50) .06 (.47) .81 (.50) .85 (.46) 1.49∗∗ (.28) .26∗∗ (.07) .13∗∗ (.02) −1.12∗∗ (.35) .27 (.25) .80∗ (.36) −.01 (.09) −.59 (.32) −.27 (.27) −.36∗∗ (.11)

−4.69∗∗ (.55)

−4.51∗∗ (.55)

−4.65∗∗ (.66) −.79 (1.30) .42 (1.03) 2.09 (1.40)

−4.88∗∗ (.63) −1.38 (1.75) .45 (1.13) 2.71 (1.95) (Continued)

73


Table 2 (Continued) Variable Proportion African American Proportion Hispanic Proportion college graduates Structural disadvantage West Midwest Northeast Cross Level Interactions with Religious Attendance Church adherents (per capita) Religious heterogeneity Conservative Protestant adherents (per capita) Proportion black Proportion Hispanic Proportion college graduates Structural disadvantage West Midwest Northeast Cross Level Interactions with Religiosity Church adherents (per capita) Religious heterogeneity Conservative Protestant adherents (per capita)

1

2

3

4

−.01 (.01) .01 (.01) .02 (.02) .04 (.13) .21 (.52) .18 (.54) .37 (.53)

−.003 (.02) .02 (.02) .03 (.02) −.02 (.14) .16 (.55) .23 (.52) .65 (1.03) ∗

−2.55 (1.08) −.48 (.65) 1.88∗ (.92) .01 (.01) .01 (.01) −.01 (.01) −.26∗∗ (.09) −.79∗∗ (.29) −.38 (.27) .09 (.55)

2.60 (1.60) .68 (1.43) −1.23 (1.63) (Continued)

74


Table 2 (Continued) Variable

1

2

3

4

19628.63b 29

−.01 (.01) .01 (.02) .03 (.03) .28 (.17) .81 (.66) .67 (.62) .26 (.72) 19575c 54

Proportion black Proportion Hispanic Proportion college graduates Structural disadvantage West Midwest Northeast Deviance statistic n parameters

19641.79 18

19638.56a 19

∗ p.

< .05; ∗∗ p. < .01 (two-tailed tests). Note: Nonaffiliated respondents are the default category for religious affiliation. a Likelihood ratio test compared to model in column 1, χ 2 = 3.11 with 1 d.f., p. = .07. b Likelihood ratio test compared to model in column 2, χ 2 = 10.03 with 10 d.f., p. = .44. c Likelihood ratio test compared to model in column 2, χ 2 = 63.48 with 35 d.f., p. = .003.

Model 2 of Table 2 includes both the individual-level variables and an interaction term, capturing the potential moderating role of religious attendance for Catholics.13 Although being Catholic is associated with an increased likelihood of reporting gambling problems, religious attendance serves to moderate this association, with greater church attendance serving to ameliorate the association between being Catholic and reporting gambling problems. Hence, a one unit increase in religious attendance for Catholic respondents is associated with approximately a 40.5 percent decrease in the expected number of gambling problems, compared to Catholic respondents reporting lower religious attendance (holding all other variables constant). While this finding is not supportive of the aforementioned hypothesis (no. 3) positing that religious attendance will be more important for predicting deviant behavior for Protestant denominations than other religious denominations, it does demonstrate the importance in evaluating the protective influence that religious attendance may exhibit across different denominations (Jensen and Erickson 1979). Model 3 of Table 2 includes the contextual-level factors in addition to the individuallevel predictors. In short, the inclusion of the county-level measures appears to add little to the prediction of reported gambling problems, with the pattern of significant predictors among the individual-level factors remaining unchanged and none of the contextual-level factors reaching statistical significance. Indeed, the likelihood ratio test comparing the results of this model with those reported in column 2 confirms the obvious—the inclusion of

13

An additional model (available upon request) that included interactions for each of the five categories of religious affiliation with the measure of religious attendance revealed that only the interaction between being Catholic and religious attendance was statistically significant. Based on this finding, the model presented assumes that the slope of religious affiliation is only different for Catholics and that all of the other categories of religious affiliation have parallel slopes for religious attendance.


75

county-level measures of religious adherence and dispersion (along with other control measures) contributes little to the prediction of gambling problems among the respondents χ 2 = 10.03 with 10 d.f., p = .44. This null finding fails to support either of the contextual-level hypotheses (the moral communities hypothesis [no. 4] and the conservative Protestantism hypothesis [no. 6]). Furthermore, results of additional analyses that excluded those individuals who reported never gambling also largely mirrored the presented analyses (results available upon request). Although such contextual measures of religious participation and diversity may not directly influence gambling problems, it is possible that such contexts moderate the association between religiosity, religious attendance, and the dependent variable (hypotheses nos. 5, 7, and 8). Model 4 of Table 2 considers possible cross-level interactions between the respondent’s religious attendance and the county-level measures, between the respondent’s religiosity and the county-level measures, and the influence of such contextual factors on the average level of reported gambling problems (i.e., the intercept as outcome model, as reported in model 2 of Table 2).14 This model examines whether the association between these measures of religion (and the dependent variable) change significantly across scores on a level 2 (county) variable. While none of the county-level measures were found to be predictive of variation in the random slope for religiosity, the model reveals a number of statistically significant interactions between county-level measures and the slope for religious attendance. Both measures of church adherents per capita— the general measure as well as the measure of conservative church adherents per capita—were found to predict variation in the religious attendance slope, but in opposite directions. First, in counties that are relatively high in church adherents, religious attendance is negatively associated with self-reported gambling problems. Hence, these results suggest that while religious attendance may not be associated with a decreased likelihood of reporting gambling problems overall, under conditions of a strong religious county (defined as a relatively greater number of church adherents per capita) a protective relationship is revealed, consistent with the moral communities hypothesis (no. 5). Figure 1 illustrates the conditional association between religious attendance and reported gambling problems. In counties with relatively few church adherents per capita or even an average number of church adherents per capita, variation in respondent church attendance has little predictive utility for the dependent variable. However, in counties with relatively high (+1 standard deviation above the mean) levels of church adherents per capita, religious attendance demonstrates a clear influence on self-reported gambling problems, with greater church attendance associated with significantly fewer reported gambling problems. However, much of this contextual effect is mitigated if most of the church adherents are conservative Protestants. Since the measure of church adherents per capita must increase as the measure of conservative Protestant church adherents per capita increases, the effect of church adherents per capita on the slope of religious attendance is greatly attenuated by the countervailing influence of conservative Protestant church adherents per capita. These results are supportive of Beyerlein and Hipp’s (2005) thesis (hypothesis no. 8) that while the extent of conservative Protestantism may do little to promote the type of capital that serves the entire area in reducing social problems, church adherents, such as mainline Protestant and Catholic, serve to promote bridging capital that increases social capital throughout the area that can serve to reduce social problems. Finally, two other contextual influences were found to predict variation in the religious attendance slope: the index of structural disadvantage and residing in a western county (relative to a southern county). The positive association between religious attendance and gambling problems was attenuated by both increased structural disadvantage and residing in a western county (relative

14

An unreported model that treated both religious attendance and religiosity as random effects revealed that there was significant variation in the slope of both measures across counties (p < .05).

76


Figure 1 Expected count for various values of religious attendance and county-level church adherents

Note: Expected count is for an unemployed Catholic male, approximately 25 years old, reporting mean levels of education, religiosity, and deviance and residing in a Southern county. All contextual factors (except church adherents per capita) are set to their respective means.

to the South). Prior research has suggested that in disadvantaged communities, churches (and religion) may be one of the few protective institutions that operate effectively, and hence religion may be a more effective deterrent in such areas than in other areas where secular institutions operate more successfully (Jang and Johnson 2001). And while the finding that residing in a western county (vs. the South) attenuates the association between church attendance and problem gambling is more vexing to interpret, one explanation is that this finding represents regional differences in the average proximity to casinos, a factor found to predict gambling problems in past studies (French, Maclean, and Ettner 2008; Welte et al. 2004, 2007).15 However, without direct measures of gambling opportunities (including proximity to casinos), such an interpretation is tenuous and demands further examination with data that include gambling opportunity measures. DISCUSSION AND CONCLUSION Based on prior theorizing and scholarship examining the association between religion and deviance, the present study examined whether contextual-level measures of the nature and strength of the larger religious context had an association with gambling problems and whether such measures moderated the relationship between individual religiosity and gambling problems among a sample of young adults. Overall, the picture painted by these analyses suggests that religious involvement matters only in some contexts for predicting gambling problems. If one considers the measure of gambling problems evaluated in this study, the hypothesis that is best supported

15

Among the reasons for why the southern respondents (relative to the western respondents) may reside, on average, closer in proximity to casinos are the following: (1) western states are the largest states in area with no large difference between the number of commercial and Native American casinos per capita in the West versus the South (with the obvious exception of Las Vegas, Nevada, a national tourist destination); and (2) two western states—Hawaii and Utah—do not have casinos or allow any form of legal gambling. However, without knowing what counties (from the West and the South) are actually represented in the present data, such inferences are simply conjectural.


77

by the results is the bonding (vs. bridging) capital thesis (hypothesis no. 8) of Beyerlein and Hipp (2005): the negative association between religious attendance (at the individual level) and problem gambling is strongest when church adherents per capita is relatively high (such a finding is also consistent with the moral communities hypothesis [no. 5]), but when conservative Protestants per capita is relatively high, religious attendance (at the individual level) is associated with an increased risk of self-reported gambling problems. While the study does not provide for any direct test of the bonding/bridging capital thesis, this pattern of findings is consistent with the notion that while greater overall church adherents per capita serves to control deviant behaviors, high rates of conservative Protestantism may serve to exacerbate such problems. This is an important finding, since prior research on gambling (measured as opposition to state-supported lotteries) found that persons who have greater church attendance are more likely to oppose gambling (Ellison and Nybroten 1999) and less likely to report gambling problems (Hoffmann 2000). The present research illustrates the importance of the influence of the larger religious context in understanding the link between church attendance and gambling. So, why did the models fail to adduce support for the notion that conservative Protestants, either individually or as a contextual influence, were at a decreased risk of self-reported gambling problems? While speculative, one answer may be that gambling is not such condemned behavior by conservative Protestants anymore. As Wolfe (2009:380) notes, “one reason that the conservative Protestants did not make gambling into a major culture war issue is that, despite their reputation as political extremists, conservative religious leaders in the United States operate as political pragmatists unwilling to criticize positions or practices popular among their members.” Indeed, Wolfe also notes that gambling is big business now in states where conservative Protestantism is strong (2009:381). However, one cannot rule out an explanation that focuses on a limitation of the data. Because the gambling measure is comprised of lifetime measures (i.e., asking if one had ever experienced such a problem), it is clearly possible that a temporal order issue is salient. For example, some young people who have experienced problems with gambling may turn to religion as a solution to their problems, especially in communities wherein there is a strong conservative Protestant presence. Furthermore, research by Blanchard and colleagues (Blanchard et al. 2008) suggests that distinctions between evangelical, fundamentalist, and Pentecostal church adherents at the county level may be important when examining the influence of such conservative Protestant religious contexts on behavior. Unfortunately, the present data do not allow for considering such distinctions. Clearly, additional research with data that can consider both the temporal order issue that the present analysis cannot untangle and distinctions among the conservative Protestants at the aggregate level is needed. The findings adduced from the current study also imply that the association between individual religious involvement/religiosity and gambling problems is more nuanced and complex than early explanations, such as aforementioned anti-asceticism hypothesis of Burkett and White (1974) suggest. Indeed, in a prior analysis using Add Health data, Regnerus found that the individual and contextual measures of religion were more strongly associated with theft than minor delinquency, leading him to conclude that “in contrast to the majority of previous research . . . religion may not be most suited to shape minor forms of delinquency as once thought” (Regnerus 2003:542; see also Nonnemaker, McNeely, and Blum 2003). While the present study does not support the notion that religion does not matter at all in predicting gambling problems, it does suggest that religion may not play the same direct role in reducing gambling behavior as it does with other forms of deviant activity. Future studies of gambling behavior (if possible) may be served by incorporating measures of religious participation because of the role these larger religious contexts may play in either attenuating or amplifying gaming activities. In this regard, the present study reaffirms the importance of considering the social structures that individuals are embedded in, and the importance of considering religion as a key part of those social structures. In considering these results, a couple of limitations of the Add Health data must be acknowledged, beyond the aforementioned concerns regarding temporal order and the measurement of

78


the dependent variable. First, it should be recognized that since this is a sample of young adults, the findings may not be applicable to other age groups. And as mentioned earlier, there were no measures of opportunities to gamble considered in these analyses. It is likely that some of the pathways in which the larger religious context influence individual gambling behaviors include controlling or reducing opportunities to engage in legally sanctioned gaming acts, such as lotteries and casinos (although the opportunity to gamble has changed dramatically in the past given the aforementioned information regarding the almost ubiquitous opportunities to gamble via lotteries, not to mention the growth of Internet gambling). Prior research has demonstrated that proximity to casinos and other legitimate gaming opportunities is associated with problem gambling (French, Maclean, and Ettner 2008; Welte et al. 2004, 2007; see also Sevigny et al. 2008, who find an inverse association between casino proximity and gambling problems).16 Further research is needed that can consider the association between county-level religious participation and legal opportunities to gamble (especially casino proximity) when evaluating the role of context in predicting individual gambling acts. Despite these limitations, these findings contribute to extant research that has explored the influence of religion on deviant behavior. The central finding, that religious involvement appears to provide a protective role in strong religious contexts, may have implications beyond simply decreasing the risk of developing gambling problems. A number of studies have found that problem and pathological gambling is comorbid with substance use disorders, suicidal ideation, and mood and anxiety disorders (e.g., Cunningham-Williams et al. 1998; el-Guebaly et al. 2006; Lesieur and Anderson 1995; Thompson, Gazel, and Rickman 1996; Petry, Stinson, and Grant 2005). Given that a number of studies have found a fairly robust link between religious involvement and better physical and mental health status (e.g., Ellison and Levin 1998; McCullough et al. 2000; Koenig, McCullough, and Larson 2001; Smith, McCullough, and Poll 2003), including serving as a protective factor against substance use disorders (e.g., Bahr et al. 1998; Bjarnason et al. 2005; Jang and Johnson 2001; Nonnemaker, McNeely, and Blum 2003), the possibility that religious involvement, especially in strong religious contexts, may serve as a protective factor against myriad problem behaviors deserves further exploration. At the least, future research that examines the association between individual and contextual religious involvement and comorbid substance use and problem/pathological gambling is warranted. REFERENCES Aasved, Mikal. 2003. The sociology of gambling. Springfield, IL: Charles C. Thomas. Abbott, Max W. and Rachel A. Volberg. 2000. Taking the pulse on gambling and problem gambling in New Zealand: Phase one of the 1999 National Prevalence Survey (report number three of the New Zealand Gaming Survey). Wellington, New Zealand: Department of Internal Affairs. Adamcyzk, Amy and Jacob Felson. 2008. Fetal positions: Unraveling the influence of religion on premarital pregnancy resolution. Social Science Quarterly 89(1):17–38. Agnew, Robert. 1992. Foundation for a general strain theory of crime and delinquency. Criminology 30(1):47–88. American Gaming Association. 2007. Industry information fact sheets: Statistics. Gaming revenue: Current-year data. Available at http://www.americangaming.org/Industry/factsheets/statistics_detail.cfv?id=7, accessed January 5, 2009. American Psychiatric Association. 1994. Diagnostic and statistical manual of mental disorders, 4th ed. Washington, DC: Author. Bahr, Stephen J., Suzanne L. Maughan, Anastasios C. Marcos, and Bingdao Li. 1998. Family, religiosity, and the risk of adolescent drug use. Journal of Marriage and the Family 60(4):979–92. Baier, Colin J. and Bradley R. E. Wright. 2001. If you love me, keep my commandments: A meta-analysis of the effect of religion on crime. Journal of Research on Crime and Delinquency 38(1):3–21.

16

Due to disclosure concerns, geocodes (which would allow one to match county or state gambling opportunity measures to the existing data) are not available in the restricted-use Add Health data.


79

Bainbridge, William Sims. 1989. The religious ecology of deviance. American Sociological Review 54(2):288–95. Bearman, Peter S., Jo Jones, and J. Richard Udry. 1997. The National Longitudinal Study of Adolescent Health: Research design. Chapel Hill, NC: Carolina Population Center. Available at http://www.copc.unc.edu/projects/ addhealth/design.html, accessed January 5, 2009. Bellair, Paul E., Vincent J. Roscigno, and Thomas L. McNulty. 2003. Linking local labor market opportunity to violent adolescent delinquency. Journal of Research in Crime and Delinquency 40(1):6–33. Benda, Brent B. 1995. The effect of religion on adolescent delinquency revisited. Journal of Research in Crime and Delinquency 32(4):446–66. Beyerlein, Kraig and John R. Hipp. 2005. Social capital, too much of a good thing? American religious traditions and community crime. Social Forces 84(2):995–1013. Bjarnason, Thoroddur, Thorolfur Thorlindsson, Inga Dora Sigfusdottir, and Michael R. Welch. 2005. Familial and religious influences on adolescent alcohol use: A multi-level study of students and school communities. Social Forces 84(1):375–90. Blanchard, Troy C., John P. Bartkowski, Todd L. Matthews, and Kent R. Kerley. 2008. Faith, morality and mortality: The ecological impact of religion on population health. Social Forces 86(4):1591–1620. Bolker, Benjamin, M., Mollie E. Brooks, Connie J. Clark, Shane W. Geange, John R. Poulsen, M. Henry H. Stevens, and Jada-Simone S. White. 2008. Generalized linear mixed models: A practical guide for ecology and evolution. Trends in Ecology and Evolution 24(3):127–35. Bradley, Martin B., Norman M. Green, Dale E. Jones, Mac Lynn, and Lou McNeil. 1992. Churches and church membership in the United States, 1990. Atlanta, GA: Glenmary Research Center. Bryk, Anthony S., Stephen W. Raudenbush, and Richard T. Congdon 1996. HLM. Hierarchical linear and nonlinear modelling with the HLM/2L and HLM/3L programs. Chicago, IL: Scientific Software International. Burdette, Amy M., Victor Wang, Glen H. Elder, Terrence D. Hill, and Janel Benson. 2009. Serving God and country? Religious involvement and military service among young adult men. Journal for the Scientific Study of Religion 48(8):794–804. Burkett, Stephen R. 1993. Perceived parents’ religiosity, friends’ drinking, and hellfire: A panel study of adolescent drinking. Review of Religious Research 35(2):136–54. Burkett, Stephen R. and Mervin White. 1974. Hellfire and delinquency: Another look. Journal for the Scientific Study of Religion 13(4):455–62. Chantala, Kim. 2006. Guidelines for analyzing Add Health data. Available at http://www.cpc.unc.edu/ projects/addhealth/data/using/guides/wt-guidelines.pdf, accessed January 5, 2009. Clotfelter, Charles T. and Phillip J. Cook. 2009. The importance of a good cause: Ends and means in state lotteries. In Gambling: Mapping the American moral landscape, edited by A. Wolfe and E. C. Owens, pp. 11–38. Waco, TX: Baylor University Press. Cochran, John K. and Ronald L. Akers. 1989. Beyond hellfire: An exploration of the variable effects of religiosity on adolescent marijuana and alcohol use. Journal of Research in Crime and Delinquency 26(3):198–225. Cunningham-Williams, Renee M., Linda B. Cottler, Wilson M. Compton III, and Edward L. Spitznagel. 1998. Taking chances: Problem gamblers and mental health disorders: Results from the St Louis Epidemiological Catchment Area Study. American Journal of Public Health 88(7):1093–96. Desmond, Scott A., George Kikuchi, and Kristopher H. Morgan. 2010. Congregations and crime: Is the spatial distribution of congregations associated with neighborhood crime rates? Journal for the Scientific Study of Religion 49(1):37– 55. Diaz, Joseph D. 2000. Religion and gambling in sin-city: A statistical analysis of the relationship between religion and gambling patterns in Las Vegas residents. Social Science Journal 37(3):453–58. Dombrink, John. 2009. The unproblematic normalization of gambling in America. In Gambling: Mapping the American moral landscape, edited by A. Wolfe and E. C. Owens, pp. 193–209. Waco, TX: Baylor University Press. Edmondson, Brad. 1986. Demographics of gambling. American Demographics 8(7):105–10. El-Guebaly, Nady, Scott B. Patten, Shawn Currie, Jeanne V. A. Williams, Cynthia A. Beck, Colleen J. Maxwell, and Jian Li Wang. 2006. Epidemiological associations between gambling behavior, substance use, and mood and anxiety disorders. Journal of Gambling Studies 22(3):275–87. Ellison, Christopher G. 1991. An eye for an eye? A note on the southern subculture of violence thesis. Social Forces 69(4):1223–39. Ellison, Christopher G., Jeffrey Burr, and Patricia McCall. 1997. Religious homogeneity and metropolitan suicide rates. Social Forces 76(1):273–99. ——. 2003. The enduring puzzle of southern homicide. Homicide Studies 7(4):326–52. Ellison, Christopher G. and Jeffrey S. Levin. 1998. The religion-health connection: Evidence, theory, and future directions. Health Education and Behavior 25(6):700–20. Ellison, Christopher G. and Michael J. McFarland. 2011. Religion and gambling among U.S. adults: Exploring the role of traditions, beliefs, practices, and networks. Journal for the Scientific Study of Religion 50(1):82–102. Ellison, Christopher G. and Kathleen A. Nybroten. 1999. Conservative Protestantism and opposition to state-sponsored lotteries: Evidence from the 1997 Texas poll. Social Science Quarterly 80(2):356–69.

80


Evans, T. David, Francis T. Cullen, R. Gregory Dunaway, and Velmer S. Burton Jr. 1995. Religion and crime reexamined: The impact of religion, secular controls, and social ecology on adult criminality. Criminology 33(2):195–224. French, Michael T., Johanna Catherine Maclean, and Susan L. Ettner. 2008. Drinkers and bettors: Investigating the complementarity of alcohol consumption and problem gambling. Drug and Alcohol Dependence 96(1–2):155–64. Gerstein, Dean R., Sally Murphy, Marianna Toce, John P. Hoffmann, Amanda Palmer, Robert Johnson, Cindy Larison, Lucien Chuchro, Tracie Bruie, Lazlo Englemann, Mary Ann Hill, Henrick Harwood, Adam Tucker, and Eugene Christiansen. 1999. Gambling impact and behavior study: Report to the National Gambling Impact Study Commission. Chicago, IL: National Opinion Research Center, University of Chicago. Grinols, Earl. 2004. Gambling in America: Costs and benefits. New York: Cambridge University Press. Gupta, Rina and Jeffrey L. Derevensky. 1998. An empirical examination of Jacob’s general theory of addictions: Do adolescent gamblers fit the theory? Journal of Gambling Studies 14(1):17–49. Hill, Jonathan P. 2009. Higher education as moral community: Institutional influences on religious participation during college. Journal for the Scientific Study of Religion 48(3):515–34. Hirschi, Travis. 1969. Causes of delinquency. Berkeley, CA: University of California Press. Hirschi, Travis and Rodney Stark. 1969. Hellfire and delinquency. Social Problems 17(2):202–13. Hoffmann, John P. 2000. Religion and problem gambling in the U.S. Review of Religious Research 41(4):488–509. Hoffmann, John P. and Stephen J. Bahr. 2006. Crime/deviance. In Handbook of religion and social institutions, edited by H. R. Ebaugh, pp. 241–64. New York: Springer. Issacs, Neil David. 2001. You bet your life: The burdens of gambling. Lexington, KY: University of Kentucky Press. Jang, Sung Joon and Byron R. Johnson. 2001. Neighborhood disorder, individual religiosity, and adolescent use of illicit drugs: A test of multilevel hypotheses. Criminology 39(1):109–44. Jensen, Gary F. and Maynard L. Erickson. 1979. The religious factor and delinquency: Another look at the hellfire hypothesis. In The religious dimension: New directions in quantitative research, edited by R. Wuthnow, pp. 157– 77, Charlottesville, VA: Academic Press, University of Virginia. Joe, Harry. 2008. Accuracy of Laplace approximation for discrete response mixed models. Computational Statistics and Data Analysis 52(12):5066–74. Johnson, Byron R., Spencer DeLi, David B. Larson, and Michael McCullough. 2000. A systematic review of the religiosity and delinquency literature. Journal of Contemporary Criminal Justice 16(1):32–52. Koenig, Harold G., Michael E. McCullough, and David B. Larson. 2001. Handbook of religion and health. New York: Oxford University Press. Korn, David A. and Howard J. Shaffer. 1999. Gambling and the health of the public: Adopting a public health perspective. Journal of Gambling Studies, 15(4):289–365. Kreft, Ita G. G., Jan De Leeuw, and Leona S. Aiken. 1995. The effect of different forms of centering in hierarchical linear models. Multivariate Behavioral Research 30(1):1–21. Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. Structural covariates of homicide rates: Are there any invariances across time and social space? American Journal of Sociology 95(4):922–63. Langhinrichsen-Rohling, Jennifer, Paul Rohde, John R. Seeley, and Martin L. Rohling. 2004. Individual, family, and peer correlates of adolescent gambling. Journal of Gambling Studies 20(1):23–46. Lee, Matthew R. 2006. The religious institutional base and violent crime in rural areas. Journal for the Scientific Study of Religion. 45(3):309–24. Lee, Matthew R. and John P. Bartkowski. 2004. Civic participation, regional subcultures, and violence: The differential effects of secular and religious participation on adult and juvenile homicide. Homicide Studies 8(1):5–39. Lee, Valerie, E., Todd K. Chow-Hoy, David T. Burkham, Douglas Geverdt, and Becky A. Smerdon. 1998. Sector differences in high school course taking: A private or Catholic school effect? Sociology of Education 71(4):314–35. Lesieur, Henry R. and Christopher W. Anderson. 1995. Results of a survey of Gamblers Anonymous members in Illinois. Park Ridge, IL: Illinois Council on Problem and Compulsive Gambling. Lesieur, Henry R. and Sheila B. Blume. 1987. The South Oaks Gambling Screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry 144(9):1184–88. McCullough, Michael E., William T. Hoyt, David B. Larson, Harold G. Koenig, and Carl Thoresen. 2000. Religious involvement and mortality: A meta-analytic review. Health Psychology 19(3):211–22. McVeigh, Rory. 2006. Structural influences on activism and crime: Identifying the social structure of discontent. American Journal of Sociology 112(2):510–66. Melnick, R. Shep. 2009. New politics, same old vice: Gambling in the twenty-first century. In Gambling: Mapping the American moral landscape, edited by A. Wolfe and E. C. Owens, pp. 97–111. Waco, TX: Baylor University Press. Nelson Hart M. and James F. Rooney. 1982. Fire and brimstone, lager and pot: Religious involvement and substance use. Sociological Analysis 43(3):247–55. Nonnemaker, James M., Clea A. McNeely, and Robert Wm. Blum. 2003. Public and private domains of religiosity and adolescent health risk behaviors: Evidence from the National Longitudinal Study of Adolescent Health. Social Science & Medicine 57(11):2049–54. Pearce, Lisa D. and Dana L. Haynie. 2004. Intergenerational religious dynamics and adolescent delinquency. Social Forces 82(4):1553–72.


81

Petry, Nancy M., Frederick S. Stinson, and Bridget F. Grant. 2005. Comorbidity of DSM-IV pathological gambling and other psychiatric disorders: Results from the national epidemiologic survey on alcohol and related conditions. Journal of Clinical Psychiatry 66(5):564–74. Regnerus, Mark D. 2003. Moral communities and adolescent delinquency: Religious contexts and community social control. Sociological Quarterly 44(4):523–54. ——. 2006. Adolescent delinquency. In Handbook of religion and social institutions, edited by H. R. Ebaugh, pp. 265–84. New York: Springer. Rohrbaugh, John and Richard Jessor. 1975. Religiosity in youth: A personal control against deviant behavior. Journal of Personality 43(1):136–55. Rostosky, Sharon S., Brian L. Wilcox, Margaret Laurie Comer Wright, and Brandy Randall. 2004. The impact of religiosity on adolescent sexual behavior: A review of the evidence. Journal of Adolescent Research 19(6):677–97. Sevigny, Serge, Robert Ladouceur, Christian Jacques, and Michael Cantinotti. 2008. Links between casino proximity and gambling participation, expenditure, and pathology. Psychology of Addictive Behaviors 22(2):295–301. Slutske, Wendy S., Kristina M. Jackson, and Kenneth J. Sher. 2003. The natural history of problem gambling from age 18 to 29. Journal of Abnormal Psychology 112(2):263–74. Smith, Christian. 2003. Theorizing religious effects among American adolescents. Journal for the Scientific Study of Religion 42(1):17–30. Smith, Timothy B., Michael E. McCullough, and Justin Poll. 2003. Religiousness and depression: Evidence for a main effect and the moderating influence of stressful life events. Psychological Bulletin 129(4):614–36. Stark, Rodney. 1996. Religion as context: Hellfire and delinquency one more time. Sociology of Religion 57(2):163–73. Stark, Rodney and William Sims Bainbridge. 1997. Religion, deviance, and social control. New York: Routledge. Stark, Rodney, Daniel P. Doyle, and Lori Kent. 1980. Rediscovering moral communities: Church membership and crime. In Understanding crime, edited by T. Hirschi and M. Gottfredson, pp. 43–52. Beverly Hills, CA: Sage. Stark, Rodney, Lori Kent, and Daniel P. Doyle. 1982. Religion and delinquency: The ecology of a “lost” relationship. Journal of Research in Crime and Delinquency 19(1):4–24. Steensland, Brian, Jerry Z. Park, Mark D. Regnerus, Lynn D. Robinson, W. Bradford Wilcox, and Robert Woodberry. 2000. The measure of American religion: Toward improving the state of the art. Social Forces 79(1):291–318. Sutherland, Edwin. 1947. Principles of criminology, 4th ed. Philadelphia, PA: J.B. Lippincott. Svetieva, Elena and Michael Walker. 2008. Inconsistency between concept and measurement: The Canadian Problem Gambling Index (CPGI). Journal of Gambling Issues 22(1):157–73. Thompson William N., Ricardo Gazel, and Dan Rickman. 1996. The social costs of gambling in Wisconsin. Wisconsin Policy Research Institute Report 9(6):1–44. Umberson, Debra. 1987. Family status and health behaviors: Social control as a dimension of social integration. Journal of Health and Social Behavior 28(3):306–19. Welte, John W., Grace M. Barnes, Marie-Cecile Tidwell, and Joseph H. Hoffman. 2008. The prevalence of problem gambling among U.S. adolescents and young adults: Results from a national survey. Journal of Gambling Studies 24(2):119–33. Welte, John W., Grace M. Barnes, William F. Wieczorek, Marie-Cecile Tidwell, and Joseph H. Hoffman. 2007. Type of gambling and availability as risk factors for problem gambling: A Tobit regression analysis by age and gender. International Gambling Studies 7(2):183–98. Welte, John W., Grace M. Barnes, William F. Wieczorek, Marie-Cecile Tidwell, and John Parker. 2002. Gambling participation in the U.S.—Results from a national survey. Journal of Gambling Studies 18(4):313–37. Welte, John W., William F. Wieczorek, Grace M. Barnes, Marie-Cecile Tidwell, and Joseph H. Hoffman. 2004. The relationship of ecological and geographic factors to gambling behavior and pathology. Journal of Gambling Studies 20(4):405–23. Wolfe, Alan. 2009. The cultural war issue that never was: Why the right and left have overlooked gambling. In Gambling: Mapping the American moral landscape, edited by A. Wolfe and E. C. Owens, pp. 373–93. Waco, TX: Baylor University Press. Wolfe, Alan and Erik C. Owens. 2009. Introduction. In Gambling: Mapping the American moral landscape, edited by A. Wolfe and E. C. Owens, pp. 1–7. Waco, TX: Baylor University Press.