KEYWORDS: power-control theory; family structure; patriarchy; deviance; public policy.

Developed by Hagan, Gillis and Simpson (Hagan 1989; Hagan, Gillis and Simpson 1985; Hagan, Simpson and Gillis 1987; Hagan, Gillis and Simpson1990), power-control theory combines class and control theories of deviance to explain the effects of familial control on gender differences in crime. Hagan et al. (1987) argue that parental positions in the workforce affect patriarchal attitudes in the household. Patriarchal attitudes, in turn, result in different levels of control placed on boys and girls in these households. Finally, differing levels of control affect the likelihood of the children taking risks and ultimately engaging in deviance. In other words, because of the greater levels of control placed on girls in patriarchal households, there are greater gender differences in delinquency in such households in that boys are more delinquent than girls.

Power-control theory begins with the assumption that mothers constitute the primary agents of socialization in the family. In households in which the mother and father have relatively similar levels of power at work, "balanced households," mothers will be less likely to differentially exert control upon their daughters. Thus, in balanced households, both sons and daughters will have similar levels of control placed upon them, leading them to develop similar attitudes regarding the risks and benefits of engaging in deviant behavior. This line of reasoning suggests that balanced households will experience fewer gender differences in deviant behavior. Power-control theorists further assume that households in which mothers and fathers have dissimilar levels of power in the work place, so-called "unbalanced households," are more "patriarchal" in their attitudes regarding gender roles. In such households parents will place greater levels of control upon daughters than upon sons. Therefore, daughters will develop attitudes unfavorable towards deviance--higher levels of perceived risk and fewer perceived benefits for engaging in deviant acts. Thus, in unbalanced households the theory predicts significant gender differences in deviant behavior, with male children being more likely than females to engage in deviant acts.

Initial tests of power-control theory suggested that these gender differences in crime come about because girls are differentially controlled in the household. In other words, female delinquency increases or decreases depending on the level of patriarchy and, thus, control in the household. Later tests of the theory (McCarthy, Hagan, and Woodward 1999) suggest that gender differences in delinquency and crime probably decrease because both male and female delinquents are affected. Most importantly, McCarthy et al. (1999) demonstrate that in less patriarchal households sons have more controls placed on them, decreasing their level of delinquency.

Most tests of power-control theory have provided at least moderate support for its model of the causes of gender-differences in deviant behavior (Hagan et al. 1985; Hagan et al. 1987; McCarthy and Hagan 1987; Singer and Levine 1988; Hagan 1989; Hagan 1990; Hagan et al. 1990; Hagan and Kay 1990; Sacco 1990; Morash and Chesney-Lind 1991; Hagan and Wheaton 1993; Jensen 1993; Grasmick et al. 1996; Heimer 1996; Avakame 1997; Leiber and Wacker 1997; McCarthy et al. 1999; Blackwell 2000; Blackwell, Sellers, and Schlaupitz 2002)-see Jensen and Thompson (1990) for an exception. As the theory predicts, gender has been found to have very little, if any, relationship to delinquency in less patriarchal households (Hagan et al. 1985, 1987, 1990). In addition to testing power-control theory, researchers have attempted to extend it in several ways. For example, theorists have modified or extended power-control theory to apply to adult populations (Grasmick et al. 1996; Blackwell 2000), to include the effects of peer relationships (Singer and Levine 1988; Avakame 1997) and religion (Avakame 1997), to predict the likelihood of victimization (Hagan 1990; Sacco 1990; Blackwell et al. 2002) and adolescent role exits (Hagan 1990; Hagan and Wheaton 1993; Blackwell et al. 2002).

This paper presents a further attempt to extend power-control theory with the inclusion of multiple types of family structure. Thus far, tests of power-control theory have focused primarily upon intact, two-parent families and single mother families (for an exception see Morash and Chesney-Lind 1991; Leiber and Wacker 1997; and Blackwell et al. 2002). While alternative family structures, such as single father families and stepfamilies, are certainly less common than two-parent and single mother families, there is no compelling reason why the theoretical arguments outlined in power-control theory must be restricted to certain types of families. In fact, by drawing upon the sociology of the family literature and arguments by Hagan et al. (1990) and Grasmick et al. (1996), a strong argument can be made to include a large variety of family forms.

While researchers have called for an expansion of power-control theory to cover different family structures (Hagan et al 1990; Blackwell 2000), such an expansion raises theoretical complications (Hagan et al. 1990; Hagan and Kay 1990; Hagan et al. 1993). According to power-control theory, a key variable in explaining gender differences in deviant behavior is the level of patriarchy within the family. However, the measure of patriarchy used by most previous studies of power-control theory is implicitly tied to family structure, due to its focus upon differential levels of power between husbands and wives (Hagan et al. 1987). Households in which husbands and wives share equal levels of power are considered "balanced" and therefore less patriarchal. Households that experience differentials in power, "unbalanced" households, will experience higher levels of patriarchy. Basing a measure of patriarchy upon differential levels of power between spouses precludes the possibility of extending power-control theory to alternate family structures, such as single-parent households. Thus, in addition to calling for an extension of power-control theory to alternate family forms, this paper will discuss how measures of patriarchy can be disentangled from family structure.

In addition to the discussion of these theoretical issues, this paper will provide an initial test of a revised model's ability to predict gender differences in deviant behavior. The measure of family structure used in this study includes biological families, single mother families, single father families and stepfamilies. Further, the analysis uses a modified measure of patriarchy necessitated by the inclusion of alternate family forms. Finally, the policy implications for an extension of this theory will be discussed. The family is viewed as strong correlate to delinquency-this theory implicitly, then, offers suggestions for formal and informal practices to control deviant behavior.

EXTENDING POWER-CONTROL THEORY TO ALTERNATE FAMILY STRUCTURES
Since the 1960s, the divorce rate and, consequently, the number of individuals remarrying sharply increased in the Unites States (Cherlin 1992). These changes have resulted in an increase in research concerning the family and specifically family structure. For a variety of reasons, most research on the effects of family structure has been limited to single mother families. First and foremost, single mother families have been at the center of the political debate concerning the importance of family life (Dowd 1997). Furthermore, single father families and stepfamilies are far less prevalent than single mother families making the collection of data on such families difficult (Cherlin 1992). In recent years, researchers have paid more attention to single fathers (Garasky and Meyer 1996) and stepfamilies (see Ganong and Coleman 1984). However, criminologists have lagged behind in the inclusion of alternate family structures in deviance models

For example, early research examining power-control theory (Hagan et al. 1985, 1990; Hill and Atkinson 1988) focused solely on two parent families. However, more thorough analyses followed that included single mother families (Hagan et al. 1987; Jensen and Thompson 1990; Leiber and Wacker 1997; Morash and Chesney-Lind 1991). Findings from studies including single mothers have been inconclusive. Some studies find a relationship between family structure, patriarchy, and gender differences in delinquency (Hagan et al. 1987), while others find no such relationship (Jensen and Thompson 1990). Still others question the way in which patriarchy is related to family structure, by questioning how patriarchy is measured for single mother families (Morash and Chesney-Lind 1991; Leiber and Wacker 1997). Many researchers have acknowledged the theoretical complications of adding extended measures of family structure to a power-control model-these researchers suggest that addressing these complications and improving model specifications is needed before family structure can be adequately addressed by power-control theory (Hagan et al 1990; Blackwell 2000).

Power-control theory implicitly assumes that processual issues (levels of patriarchy and types of control) in the household determine gender differences in delinquency. However, while power-control theory argues that the level of power parents bring to the relationship determine the control placed upon juveniles, it is also plausible to argue that family structure itself may have significant effects upon levels of power in the household and the type and level of control exerted upon juveniles. For example, children from a single parent or stepfamily may have lower levels of control placed on them than children from biological families. In other words, if relational control is measured as the extent to which a child will want to be like his/her mother or father, this child may report a weaker connection to a non-custodial or stepparent than a custodial or biological parent. In addition, research has shown that children from stepfamilies often experience low levels of control or discipline as stepparents adjust to the parenting role (Papernow 1992). Given recent increases in single father families (Garasky and Meyer 1996) and stepfamilies (Cherlin 1992), an expansion of power-control theory to such family structures is in order.

FAMILY STRUCTURE AND THE MEASUREMENT OF PATRIARCHY
As discussed above, the inclusion of family forms beyond intact, two-parent households to power-control theory is problematic, given the theory's focus upon differentials in power between spouses in a household. Therefore, the inclusion of alternative family structures in power-control analyses requires the researcher to address the issue of how patriarchy will be measured. After all, in a single-parent household there is no spouse against whom to compare levels of power. While single father families and stepfamilies have been ignored in tests of power-control theory, some studies have included single mother families in analyses (Hagan et al. 1987; Jensen and Thompson 1990; Leiber and Wacker 1997; Morash and Chesney-Lind 1991). Therefore, a discussion of how tests of power-control theory that include single mother families measure patriarchy is in order.

Most power-control research has analyzed less patriarchal households and more patriarchal households separately (Hagan et al. 1985; Hagan et al. 1987; Jensen and Thompson 1990; Singer and Levine 1988; Grasmick et al. 1996). Those few studies that consider non-intact families have compared female-headed households to balanced and unbalanced households (Hagan et al. 1987; Jensen and Thompson 1990). The implicit assumption behind such studies is that female-headed families are a special example of a balanced household, since there is no "power imbalance" in the household (Hagan et al. 1987). Rather than examining the effect of family structure on patriarchy, female-headed households have simply become a proxy measure for low patriarchy. However, others argue that such a classification overlooks structural inequalities that women might experience in the workforce, thereby ignoring a central tenet of power-control theory (Morash and Chesney-Lind 1991). Such researchers do not assume that female-headed houses are balanced, but classify them according to the level of power the woman holds in her job (Morash and Chesney-Lind 1991; see also Lieber and Wacker 1997). While a single mother may not have to contend with power differentials with her husband, the type of job she holds may still affect her parenting style. Prior research has not examined the relationship between family structure (i.e., single mother households) and levels of patriarchy; instead it has examined the effect of single mother families on the level of support and control placed on sons and daughters. This blanket assumption that single mother families are less patriarchal in nature ignores both the complexities of single parenting and the structural position of women in society.

Single mothers (and fathers for that matter) must contend with a variety of power struggles in their daily lives, such as the continued existence of the non-custodial parent in child-rearing and decision making [i.e., single parents who must rely on child-support payments or are limited in decision-making power because the non-custodial parent has partial custody], the potential involvement of the state [i.e., single mothers or single fathers who are jobless and must rely on the state for monies], as well as power differentials at work. Hagan et al., (1990) and most researchers have focused on patriarchy in the household between custodial parents, assigning single mothers the role of less patriarchal households. Morash and Chesney-Lind (1991) have extended this measure of patriarchy to include power differentials at work, which allows single mothers the possibility of being either patriarchal or non-patriarchal. Neither has extended the argument to other non-traditional family forms, nor allowed for an argument that power differentials for single parents may exist beyond the custodial household or workplace.1

Thus, power-control research, to date, has measured patriarchy in one of two ways--both structural in nature. The more traditional measure, specified by Hagan et al (1990), is relational in nature. Relational measures determine levels of patriarchy by examining the differing levels of power conferred upon spouses by their occupations. For example, if a husband works in a job where he manages others and his wife does not, that household will be defined as patriarchal under a relational measure.

Morash and Chesney-Lind (1990) provide a different measure of patriarchy that is also structurally-based. This measure of patriarchy, perhaps best labeled a global measure, takes into account the fact that woman generally hold less power in the workforce. Women who would have no power differential in the household (for example, single mothers) may still experience a power differential in the workplace, and this general or global power differential may translate into increased control and supervision for daughters. This global measure, therefore, is a measure of female power in the workforce. For example, a household would be defined as patriarchal under the global measure if the female in the household held relatively little to no power in the workforce whether or not her partner was in a powerful position. A mother who managed a workgroup would be in a non-patriarchal household, while a mother who had relatively little power in her job would come from a patriarchal household when using a global measure of patriarchy.

While this global measure may not be the original measure specified by power-control theory, it is an important extension of the theory, especially if two parents are not present in the household. However, both of these measures, relational and global, are based upon structural characteristics of the individual or family.

AN ALTERNATIVE MEASURE OF PATRIARCHY
An alternative to structural measures of patriarchy is an attitudinal measure of patriarchy, although this has rarely been used in power-control research. Hagan et al. (1990) do suggest that a measure of marital power [as seen in Bloode and Wolfe (1960)] and decision making might provide a useful extension of structural measures of patriarchy. Furthermore, Grasmick et al. (1996) compared the effectiveness of three measures of patriarchy in predicting preferences for risk, finding that a patriarchy measure that uses both structural and attitudinal indicators was the best fit with other theoretically relevant constructs. However, with these notable exceptions, most research has remained faithful to a structural measure of patriarchy.

Family research, on the other hand, consistently measures patriarchy attitudinally, as sex role attitudes in the family. Extensive research examines the effects of family structure, specifically single mother families, on attitudes in the family (Richmond-Abbott 1984; Finlay, Starnes, and Alvarez 1985; Wright and Young 1998; Slavkin and Stright 2000; Slavkin 2001). Most relevant to the present study is research examining the effect of family structure on sex-role attitudes of parents and children. While such research has typically focused on the attitudes of either single mothers, or children from single mother households (Demo and Acock 1988), more recent research has found differences in gender role attitudes between intact households and single parent households in general (Amato and Booth 1991; Slavkin 2001). This body of research supports Hagan et al's (1990) earlier claims that an attitudinal measure of patriarchy may be useful.

For several decades, family research has acknowledged the importance of family structure on the effect of attitudes and behavior in the household:

A single parent will of necessity have to do a large number of behaviors typically stereotyped as both masculine and feminine. He or she may well come to believe in the appropriateness of less sex-role behavior and may serve as a model of more non-traditional behavior to children in the family. The children in turn may internalize their parents more flexible beliefs about appropriate sex-role behaviors and behave in a way that reflects their possession of more contemporary values. Thus, single parent families may be incubators for more sex-role values and behaviors (Richmond-Abbott 1984: 61-62).

Research examining the effect of family structure on parental attitudes has found that single mothers are, in general, more liberal in their sex-role attitudes than married women (Finlay et al. 1985), and that both sons and daughters echo their mother's attitudes (Amato and Booth 1991). While single mothers and fathers take on a wider range of responsibilities in the family and thus become more non-traditional role models (Slavkin 2001) with less restrictive rules than married parents (Thompson, McLanahan, and Curtin 1992), some research has found that children from single father families have more traditional sex role attitudes than children from single mother families (Wright and Young 1998). However, other research has found that both single mothers and single fathers have less traditional sex role behaviors and attitudes than two parent households and that these less traditional attitudes are passed on to both their sons and daughters (Richmond-Abbott 1984).

THE CURRENT THEORY AND RESEARCH
Since power-control theory provides a rather complex theoretical model, the authors are providing this model-with the inclusion of family structure-in Figure 1. Power-control theory argues that patriarchal attitudes affect delinquent behavior via the differential controls placed upon children by family members (Hagan et al. 1990). In particular, the theory argues that girls are subject to greater control than boys in patriarchal households. This study includes measures of two different forms of control for each parent--instrumental and relational (see Hagan et al. 1990). Initial specifications of the theory focused solely on maternal control, arguing that different levels of patriarchy will differentially affect how women parent. Later applications of the theory (Grasmick et al. 1996; Blackwell 2000) have included measures of paternal control, too.

In turn, consistent with power-control theory, instrumental and relational control affect levels of risk by making it more likely that girls will perceive they will get in trouble for their bad behavior and less likely to prefer to engage in risky behavior. These differentials in risk preference and perceived risk will then lead to differences in levels of deviance.

The literature suggests that the relationship between family structure and power-control processes is sufficiently complicated to warrant a more detailed analysis than traditional power-control models. This study examines the effects of family structure on patriarchy instead of using structure as a proxy for patriarchy in the household. Furthermore, rather than simply comparing single mother households to intact biological households, single fathers and stepfamilies are included in the analyses. Because this paper makes modifications to power-control theory in order to include such alternate family forms, an interaction term between gender and patriarchal attitudes was included in the model to test the assertion of the theory that patriarchal attitudes exerted on females impacts their level of deviance. Thus, this study analyzes the effect of family structure, gender, race, and age on patriarchal attitudes, levels of control, perceived risk, and, finally, deviance.

HYPOTHESES
The above discussion suggests several hypotheses. First, family structure is not expected to have a direct effect on deviance. Rather family structure should have an indirect effect through the mechanisms of patriarchy, specifically patriarchal attitudes. As previous research shows, sex-role attitudes in single parent families may become less traditional as these parents engage in both maternal and paternal activities. Single mother and single father families should have lower levels of patriarchy than biological intact families. Given that this earlier research suggests it is the absence of two parents that will lead to changes in patriarchal attitudes in the household, as single parents take on the daily roles and responsibilities of both parents, it is expected that there will be no difference in patriarchal attitudes between stepfamilies and biological intact families (since both are two parent households).

Furthermore patriarchal households should experience higher levels of control placed on daughters. Higher levels of control will, in turn, lead to a reduced preference for risky behavior and a higher sense of perceived risk for engaging in delinquent acts. Low levels of perceived risk and the preference for risky activities should have significant effects on deviant behavior, which means that in patriarchal households there should be a greater gender difference in deviance than in less patriarchal households. Put another way, gender should not have a direct effect on deviance, as its effects should be moderated by the differential controls placed upon boys and girls.

METHODS

Data
As discussed at length above, this study extends power-control theory to include family types beyond two-parent, biological families. Specifically, the analysis includes single father families, single mother families and stepfamilies. The inclusion of these alternate family structures required the development of an attitudinal measure of patriarchy, since tests of power-control theory have traditionally relied upon measures intrinsically tied to family structure. Thus, the analysis suggested by these modifications to power-control theory requires a data source that includes several items. First, the data must include sufficient information about the respondent's family to allow analyses to differentiate among intact biological families, stepfamilies, single father households, and single mother households. The data must also allow the development of both a structural and attitudinal measure of patriarchy. Further, the data must include items that allow the measurement of control within the household, since power-control theory argues that patriarchy determines the levels of control placed upon youth. Finally, in addition to basic demographic information, the data must also include measures of the perceived risks of engaging in deviance and the extent to which the respondent has actually engaged in deviant acts.

Click to view Figure 1.

Unfortunately, no existing data set that the authors could locate provided all of the necessary pieces for a test of the expanded power-control theory outlined in the above discussion. Therefore, the authors developed and administered a survey that would allow a preliminary test of this model. In the fall of 1997 and the summer of 1998, a convenience sample of 700 students in an introductory sociology course at a large, public university on the West coast was administered anonymous, multiple choice surveys. Questions consisted of demographic items (including questions regarding family structure), questions about deviant behavior, and items used to construct measures of structural patriarchy, patriarchal attitudes, instrumental and relational control, perceived risks of committing deviant acts, and the preference for risky behavior.2 Of the seven hundred questionnaires distributed, 555 were returned complete and coded for analysis. When cases with missing data on key variables were eliminated, the final number of cases in the analysis dropped to 534.

As with all studies based on college samples, there are concerns as to how representative the sample is to the general population. In fact, a college sample will differ in many ways from the general population. Such samples will tend to be disproportionately female, have a higher level of income, and have a lower overall level of deviance. As a consequence, this study cannot be assumed to apply to the general population. For studies focusing on the relationship between levels of income and patriarchal attitudes, a college sample would not provide sufficient variance in levels of income. Furthermore a college sample would clearly be inappropriate for studying predictors of serious forms of deviance. However, power-control theory focuses upon gender differences in lesser or minor forms of deviance (Hagan et al. 1985)-forms of deviance for which there is sufficient variance, even in a college sample, to allow for a statistical test of the theory. Given this paper's focus on the methodological and theoretical implications of including alternate family forms in a power-control model, a college sample was appropriate for a preliminary test.

As expected, females are, indeed, over-represented in the sample-sixty-three percent of the sample was female, thirty-seven percent males. The majority of respondents were white (68%). The next largest racial grouping was Asian (18.7%), followed by Pacific Islander (3.8%), Hispanic (3.5%), African American (2.7%), and Native American (1%). Approximately two percent of the sample indicated that they did not consider themselves a member of any of the above racial/ethnic groups. Most respondents were in their first quarter and year of college when completing the survey. The mean age for respondents was nineteen, with eighty-seven percent of the respondents under the age of 21.

Respondents were asked if, compared to other American families, their level of income was far below average, below average, average, above average or far above average. Again, since this study utilized college students, most respondents indicate that their income was above average (41%) or average (35.3%). Thirteen percent of the sample indicated that their family income was below average, and two percent indicated an income far below average. Eight percent of the sample had an income they considered "far above average."3

Because the current study examines the effect of family structure on the mechanism of power-control theory, included among the demographic items was a question that asked respondents who they lived with at age sixteen.4 The majority of respondents (74%) lived with both of their birth parents.5 Of the remaining twenty-six percent of respondents, seven percent lived with a parent and stepparent, fifteen percent lived with their birth mother only, and four percent lived with their father only.6 Twenty-one cases that indicated that they lived with "Other" persons were also excluded from the analysis.7 Family structure was entered into the regression analyses as a series of dichotomous variables, with "Lived with Both Parents" as the contrast category. Other standard demographic variables included in the analysis were gender (1 = female, 0 = male), age, and race (1 = white, 0 = non-white).

This study was conducted using a retrospective design, first suggested by Hagan (1989) as a plausible way to test theories such as power-control which have historical components. Given the difficulties associated with collecting longitudinal data, retrospective designs serve as a reasonable alternative. There are numerous arguments against retrospective design, greatest of which is that recall deteriorates over time. However, research has shown that information important to respondents is easier and more accurately recalled than information that is less important (Blair et al. 1991). For example, studies have shown that memories of many family issues such as parental supervision (Brewin et al. 1993; Parker 1989) or parental work histories (Robins et al. 1985) remain generally stable over time. In fact, Robins et al. (1985) found that a majority of respondents could accurately recall family details over a thirty year period when compared to official family records. For a thorough and detailed discussion of the costs and benefits of this design see Blackwell (2000: 455-456, 460).

Patriarchy
The current study cannot measure patriarchy in the same manner as Hagan, Simpson and Gillis (1987), as their measure is tied to family structure. The method of classification used by Morash and Chesney-Lind (1991) is less problematic, although it does not address family structures other than single mother households. This study modified this construct to work with other non-traditional family forms. Taking the lead from sociology of the family research [Hagan et al. (1990), Grasmick et al. (1996), and Morash and Chesney-Lind (1991)] both structural and attitudinal measures of patriarchy were initially included.

Global structural patriarchy
Morash and Chesney-Lind (1991) suggest an alternative structural measure of patriarchy to the Hagan, Simpson and Gillis (1987) measure. This measure of patriarchy uses the mother's structural place in the workforce instead of her comparative place given her husband's job.

Two questions were used to determine the likelihood of global patriarchy in each household. Each respondent indicated the level of power held by his or her mother (if applicable) in her occupation and the level of power held by the father (if applicable) in his occupation. For example, when indicating the occupational power held by the mother, respondents could select from "self-employed with no one else working for her," "self-employed with other people working for her," "worked for someone else and did not supervise others," "worked for someone else and supervised others," or "she was not employed." The likelihood of global patriarchy for traditional-, single mother-, and stepfamilies was based on the mother's level of power in the workforce.

Households where mothers held a position of power in the work force whether or not another adult in the household also held power were considered low patriarchy. Households where mothers did not hold a position of power in the work force were considered high patriarchy. Single father households presented a problem given Morash and Chesney-Lind's (1991) articulation of global patriarchy (which is based solely on female occupational power). In the end, global patriarchy for single father households was coded as follows: households where the father held power in the workforce were coded high patriarchy, households where the father did not hold power in the workforce were coded low patriarchy. This decision was made because previous research has determined balanced households to be those in which male power approximated levels of female power (within or outside the household). Given that women are more likely to hold jobs with little power-a balanced or less patriarchal household would be one in which the male would also hold little power in his job. Thus balanced single father households are those in which the man does not hold power in his job. Patriarchal or unbalanced households are those in which the man holds a position of power. About thirty-eight percent of households were coded as non-patriarchal, 62.2 percent of households were coded as patriarchal.

Attitudinal patriarchy
An alternative to coding patriarchy using parental occupations is to measure the level of patriarchal attitudes held by the present parent(s). After all, power-control theory argues that experiences in the workplace will affect attitudes that, in turn, will impact the amount of control exerted in the household. Numerous studies in the family literature have, in fact, measured patriarchy using attitudinal items (Finlay et al. 1985; Richmond-Abbott 1984; and Thornton et al. 1983).

A series of nine questions was used to measure the patriarchal attitudes of the respondent's father. For example, each respondent was asked if his/her father would "strongly agree," "agree," "disagree," or "strongly disagree," with the following statement: "It is much better for everyone if the man earns the main living and the woman takes care of the home and family." The remaining questions provide further indicators of the fathers' sex-role attitudes, asking about such items as child rearing, housework, and authority in the family. The respondents were then asked how their mothers would answer the same nine questions.8

The final Attitudinal Patriarchy score for each respondent was created based on his or her family structure. Separate Attitudinal Patriarchy scores were created for the father and mother by adding together their respective responses. Individual items were reverse coded as necessary, such that higher scores on the Attitudinal Patriarchy variable indicate greater patriarchal attitudes. If the respondent lived with both parents, or a parent and stepparent, their final Attitudinal Patriarchy score was the average of their parent's Attitudinal Patriarchy scores. If a respondent lived with his mother only, then his or her final Attitudinal Patriarchy score was the total score for his mother. If a respondent lived with his father only, then the final Attitudinal Patriarchy score was the father's total score. The final Attitudinal Patriarchy scores ranged from 12 to 44, with a mean of 23.73 (alpha = .75).

Instrumental and relational control
Instrumental control refers to parental awareness of their children's activities. Parents who always know what their children are doing can exert greater amounts of control over them. Respondents were asked about each parent's awareness of their activities with two questions. Respondents were asked to agree or disagree (using a four-point Likert-type scale) that the respondent's mother/father generally knew his or her whereabouts when away from home, and that the respondent's mother/father generally knew whom he or she was with when away from home. Fathers had a mean of 5.09 on the instrumental control scale. Mothers had a higher score of 6.09.

While parents can directly control the behavior of their children by monitoring their activities and friendships, they can also exert a more indirect form of control through attachment (Hirschi 1969). If a child has a strong parental relationship, he or she will not want to jeopardize that relationship by engaging in delinquent behavior. To create the measure of Relational Control respondents were asked how often they talk with their mother/father about their thoughts and feelings, and if they would like to be the "kind of person" their mother/father is. Fathers had a mean of 4.83 on the relational control index, while mothers had a mean of 5.56.

Perceived risk and risk preference
According to power-control theory, girls in patriarchal households are subject to greater levels of control. As a consequence, girls in such households do not develop a "taste" for risky beliefs or behavior. A reduced preference for risky beliefs or behavior leads, in turn, to reduced levels of delinquency. This study measures the respondent's risky attitudes using two variables-the perceived risk of engaging in risky behaviors and the respondent's preference for risky behaviors.

The perceived risk of engaging in deviant behavior was measured using three questions. Each question asked respondents whether they could engage in a particular form of delinquent behavior (breaking into a place, stealing from a store, and writing graffiti) without getting caught. Respondents answered using a four point, Likert-type scale, i.e. respondents strongly agreed, agreed, disagreed, or strongly disagreed that they could get away with the particular behavior. The perceived risk items were coded such that high scores equal higher levels of perceived risk for engaging in deviance and then they were added together. Scores ranged from three to twelve, with a mean of 7.31 and alpha value of .86.

Preference for risky activities was also measured using three questions. Respondents were asked if they strongly agree, agree, disagree, or strongly disagree with the following three item-"I like to test myself every now and then by doing something a little risky," "I sometimes find it exciting to do things for which I might get in trouble," and "Excitement and adventure are more important to me than security." These three items were coded such that higher scores indicate a higher preference for risky activities, and then they were added together. Scores for risk preference also ranged from three to twelve, with a mean of 7.99 and alpha value of .63.

Deviance
The dependent variable in the analyses, deviance, was created using four items.9 Respondents were asked if they have ever been picked up by the police, shoplifted, tried marijuana, or tried cocaine. This measure is referred to as deviance instead of delinquency because some of the respondents may have been over 18 while engaging in these acts. However, whether they were older or younger than 18, all still represent illegal acts or trouble with the law. These behaviors are in keeping with the theory's initial and continued focus on non-serious criminal behavior (Hagan et al. 1985).

A weakness in the available deviance data is the inability to place those acts of deviance within a particular time range-only prevalence measures were available. While the ability to place deviant acts for certain within the time that the respondent was living at home would be ideal, the available deviance measure provides a reasonable compromise.10 Most respondents in the sample were in their first quarter of their first year of college, increasing the likelihood that they either currently lived at home or had only recently moved away. Furthermore, power-control theory argues that differing patriarchal attitudes lead to the development of differing preferences for risk-taking behavior. As Hagan et al. (1987) and Blackwell (2000) have argued, there is no reason to expect, theoretically, that those developed risk preferences disappear once the child leaves the home.

The four available deviance items were coded as dichotomous and summed, such that higher scores equal higher levels of deviant behavior. Scores on the dependent variable ranged from zero to four, with a mean of 1.95 and alpha value of .61.

ANALYSIS AND RESULTS
While previous researchers (Hagan et al 1987, 1990; Hagan and Wheaton 1993; Blackwell 2000) have found support for the relational structural measure of patriarchy, given the inclusion of single parent households this measure could not be used in analyses. Therefore, initial analyses were run first including the global structural measure of patriarchy and second with the attitudinal measure of patriarchy. The global structural measure was not significantly correlated to control, risk, or deviance. In addition, family structure was not a predictor of global patriarchy. Thus, this measure was dropped from subsequent analyses and is not presented in the tables. The authors can only speculate as to why the global measure of patriarchy was not a significant predictor in the model. First, given the sample used in this study, a global patriarchy measure (which is based on a structural measure of class) may not be varied enough. As already reported, middle and upper-middle class respondents are over represented in this sample. If a more generalizable sample were used, global patriarchy may be significant. However, if sampling is not the problem, a non-significant finding between family structure and global patriarchy can be argued to support the original proposition that no family structure should be assumed to be a proxy for level of patriarchy (although the study did hypothesize differences, finding no difference shows the importance of separating family structure from patriarchy). Finally, the finding that global patriarchy was not significantly related to endogenous variables is in keeping with the findings from Morash and Chesney-Lind (1991) who also found in their final model that women's workplace power was not a significant predictor. While global patriarchy was not significant in this study, future research is needed and special attention should still be paid to the theoretical construction of structural patriarchy, whether relational or global, within power-control models.

The correlations between key variables in the model are supportive of power-control theory. Gender (1=female), is significantly and negatively correlated with deviance. In other words, females reported lower levels of deviant behavior than males. Also as expected the level of each parent's instrumental control (monitoring a child's behavior) is significantly and negatively correlated with deviant behavior. Curiously, the correlations for maternal relational control and paternal relational control are not consistent with the theory. Maternal relational control is not significantly correlated with deviance. On the other hand, respondents that indicated high levels of relational control from the father reported greater levels of deviance. As expected both risk variables are significantly correlated with deviance. The greater the perceived risk, the lower the levels of deviance (b = -.320, p<.01) while, the greater the preference for risk behaviors, the higher the levels of deviance (b = .353, p<.01).

Click to view Table 1.

The final analyses present a power-control model quite similar to previous tests of the theory but with the addition of extended family structure measures and an attitudinal measure of patriarchy. The model is presented as a series of regression equations. Sex (1=female), age, race (1=non-white), and a series of dichotomous variables representing family structure were included as exogenous variables. Per power-control theory, patriarchal attitudes are expected to affect the controls placed on children (instrumental and relational). Those controls, in turn, will help determine the child's preference for risky activities and his/her perception of the risk involved in deviant activities. This, in turn, will help determine the likelihood of engaging in deviance.

Table 2 presents five OLS regression equations. Patriarchy was first regressed on the demographic variables (Equation 1). Patriarchy was significantly related to family structure; however, contrary to the hypothesis, single father families tend to have higher levels of patriarchy than intact, biological families (b = 4.20, p<.01). As expected, single mothers tend to hold less patriarchal attitudes than intact, biological families (b = -4.71, p<.01). Families with stepparents, however, were not significantly different from families with both biological parents.

Click to view Table 2.

While age and race were not significantly related to patriarchal attitudes, gender did have a significant effect on perceived levels of patriarchy; females report significantly higher levels of patriarchy than males (b = .62, p<.01). This is not surprising, given research on privilege and marginalization, or what Goode (1982) refers to as the Sociology of the Superordinates which suggests that those in privileged positions (such as boys in relation to patriarchy) would be less likely to perceive this patriarchy than girls who would be more affected by this phenomenon.

The last four regressions presented in Table 2 include an interaction term between gender and patriarchy, in order to test power-control theory's central hypothesis-that the interaction between gender and patriarchy predicts the gender difference in deviance. The regressions of the measures of instrumental and relational control on the demographic variables and patriarchal attitudes do not completely support power-control theory. For example, patriarchal attitudes do not appear to affect the mother's level of instrumental control over her children (Equation 2).

However, the gender x patriarchy interaction is significant in predicting paternal instrumental control (Equation 3). The significant, negative effect of the interaction term on paternal instrumental control (b = -.30, p<.05) and the significant, positive effect of gender (b = .43, p<.05), suggest that at low conditions of patriarchy, females experience higher levels of paternal instrumental control than males (see Aiken and West 1991). But, at higher levels of patriarchy, the level of paternal instrumental control experienced by females is reduced at a steeper rate than for males. The regression of paternal relational control produced similar results Gender is significantly and positively related to paternal relational control (b = .34, p<.05), suggesting that at low levels of patriarchy, females experience higher levels of relational control. Again, the slope for males was much steeper than for females (b = -.29, p<.05), suggesting that females experience lower levels of paternal relational control at higher levels of patriarchy (b = -.29, p<.05).

While power-control theory does not specifically address the relationship between patriarchy and father's level of control and previous research has assumed that patriarchy should affect paternal control similarly to maternal control, this finding should not be unexpected. Households that experience traditional beliefs about gender roles and norms should perhaps see a lesser degree of paternal participation in child rearing. This lesser degree of parenting by fathers can translate into lower levels of paternal controls placed on children.

Gender and patriarchal attitudes also have significant effects on the relational control of mothers (Equation 4). At levels of patriarchy, females exhibit higher levels of maternal relational control than males (b = .77, p<.01). As power-control would predict, higher levels of patriarchal attitudes are associated with higher relational control of mothers (b = .09, p<.05). However the interaction term is not significant, suggesting that the relationship between gender, patriarchy, and maternal relational control is not significantly conditioned by patriarchy.

Single parent families appear to have more difficulty keeping track of their children than traditional families. Single mother families have significantly lower levels of instrumental control than intact biological families (b = -.85, p<.05) (Equation 2). Single father families have significantly lower levels of instrumental control (b = -1.37, p<.01) and relational control (b = -.52, p<.01) than intact biological families (see Equations 3 and 5).

While race does not have a significant effect on instrumental control, race does affect both maternal and paternal relational control. White respondents are the subject of significantly less maternal relational controls (b = -.34, p<.05) and paternal relational controls (b = -.52, p<.01) than non-whites. In other words, non-whites were much more likely to feel close to their parents than whites. Age had no significant effects in the regression models.

In sum, these findings suggest, consistent with power-control theory, that females are the subject of higher levels of control, both instrumental and relational, across parents. But in the case of paternal control, higher levels of patriarchal attitudes were associated with lower levels of control. These findings suggest that children may harbor negative feelings towards fathers who exhibit highly traditional gender role attitudes and may, therefore, resist their father's control.

Table 3 presents OLS regressions of perceived risk and risk preference on the power-control model variables (Equation 6 and 7). Given the non-linear distribution of the deviance measure, ordered logit was used to examine the effects of the independent variables on deviance (Equation 8). McKelvey and Zavoina (1975) first suggested that many of the dependent variables in the social sciences, while treated as latent interval level variables, were actually imperfect representations and should, instead, be treated as ordinal level variables. For example, deviance, while being scaled, cannot be assumed to have equal distance between each of its categories. Using OLS on variables of this nature may provide incorrect results (Long 1997). For this reason, the last model in Table 3 is not OLS but ordered logit instead.11

Click to view Table 3.

The family structure variables drop from significance at this stage in the model, affecting neither perceived risk, risk preference, nor deviance. Gender however, does affect both risk variables. Females perceive more risk in deviant behavior (b = 2.01, p<.01) and show less preference for risky activities (b = -.56, p<.01) than do males. Age also has a significant effect on the preference for risk-older respondents found risky behavior less attractive (b = -.06, p<.01).

The effects of patriarchy and the control variables on perceived risk and risk preference only partially support power-control theory. While paternal controls do not have a significant impact on perceived risk or risk preference, maternal instrumental control does increase the perceived risk of deviance (b = .24, p<.05).

Finally, the full model, predicting deviance, is presented in Equation 8. Maternal instrumental control also has a significant and negative effect on deviance (b = -.16, p<.01). In other words, when controlling for the preference for risk and perceived risks, the mother's awareness of her child's activities still has a negative effect on deviance. Paternal relational control has a positive effect on deviance (b = .14, p<.01), indicating that the more a student admits to wanting to be like his or her father, the more likely he or she is to have increased levels of deviance.

The risk variables have the predicted effects on deviance. As the perceived risk of deviant behavior increases, the likelihood of deviance decreases (b = -.08, p<.01). Furthermore, as the preference for risky behaviors increase, deviant behavior increases (b = .23, p<.01).

Of greatest significance is the power-control model's ability to explain the gender difference in deviant behavior. While gender does have indirect effects on deviance through the parental control variables, patriarchal attitudes, risk preference and perceived risk, gender does not have a significant effect on deviance in the final model. In order to further elaborate the various effects gender has on deviance through the variables in the power-control model, we use a standard decomposition technique in path analysis to decompose the total correlation between gender and deviance (r = -.164-see Table 1) into three components: total direct effect, indirect effects, and unmeasured/spurious effects (see Pedhauzer 1982). Standardized coefficients from the regression models are used as path coefficients. Table 4 presents a decomposition of the total gender effect on deviance.

As power-control theory predicts and Table 4 demonstrates, the majority of gender's effect on deviance is indirect (83%). The largest indirect effect of gender on deviance is via perceived risk (-.0555). Females are more likely to find deviant behavior to be a risky proposition and those who find deviance risky are less likely to engage in it. Gender also has a significant indirect effect via risk preference (-.0464). In other words, females show less of a preference for risky endeavors and those that dislike risky activities will be less likely to engage in deviance.

Gender had indirect effects on deviance through maternal instrumental control in two ways. First, gender has an indirect effect via maternal instrumental control alone (-.0228). Thus, females will be the subject of higher levels of monitoring by mothers. Those that are the subject of greater levels of instrumental control will engage in deviance with lower frequency. Maternal instrumental control and perceived risk also work together to moderate the effect of gender on deviance (-.0027) In other words, females will experience greater levels of maternal instrumental control. More control leads to a greater perception of perceived risk in deviance, which leads to lesser amounts of deviant behavior.

Click to view Table 4.

The regression models in Equations 2-8 test the proposition that patriarchy has different effects for males and females. The analysis shows that this interaction is present when predicting and paternal relational and instrumental control. Recall from our discussion of Equation 5 above that patriarchy has a negative effect on paternal relational control for both males and females, but almost twice the negative effect for females. The interaction between gender and patriarchy is linked to deviance indirectly through paternal relational control. The indirect interaction effect is negative, but quite small (-.006), and represents less than 4 percent of the total effect of gender on deviance. It suggests that patriarchy has a slight negative effect on deviance for women through paternal relational control.

Of greatest interest to the current paper are the effects of family structure on deviance. Family structure does, indeed, have important, indirect effects on deviance in a power-control model. Single fathers will tend to have greater levels of patriarchy than intact, biological families while, at the same time, have a decreased ability to monitor their children. While single mothers have decreased levels of patriarchal attitudes, they also suffer from a decreased ability to monitor children. As hypothesized, stepfamilies are not significantly different than biological families on levels of patriarchy (and, while not hypothesized, are not significantly different than biological families on their ability to monitor.) These findings suggest that, at least for patriarchal attitudes and supervision, the composition of the household is important.

CONCLUSION
This study suggests both a future direction for power-control theory and several practical directions for reducing deviant behavior. First, while Hagan et al. (1990) have suggested that attitudinal measures of patriarchy may be beneficial, research to this stage has focused on structural measures. As this study shows, this does not allow for a full examination of various family forms. Studies should begin to extend their definition of family and of patriarchy so that this theory may broaden its scope.

For this to happen, data must be collected that allows this research to move forward. In this study a convenience sample was used because existing data did not have sufficient measures on family structure, both structural and attitudinal patriarchy, instrumental and relational control, perceived risk, risk preference, and deviance. Most data sets do not give enough attention to the needed family structure variables (let alone, extended family issues, such as grandparents, siblings, non-custodial parents).

Implications for Power-Control Theory: New Theoretical Directions
While this study has respecified patriarchy in order to make room for family structure in a power-control model, additional theoretical directions may also prove useful for extending the power-control theory to include alternate family forms. Both Colvin and Pauly's (1983) integrated structural-marxist theory and Tittle's (1995) control balance theory may help extend power-control theory.

For example, integrated structural-marxist theory introduces a different way of formulating power as level of workplace control. This may be a beneficial way of discussing levels of patriarchy in that it can expand the initial formulation of power to move beyond a relational measure (which as the authors have discussed is not conducive to alternate family forms). Colvin and Pauly argue that workplace control has an effect on how parents then control their own children. An individual who has a large amount of control placed on them at work is more likely to instill in their children the importance of following the rules, whereas an individual who has less workplace control placed on them is more likely to emphasize initiative and creativity. Because this formulation does not emphasize maternal vs. paternal workplace control it is easier to see how one might include alternate family structures in this model. Maternal workplace control and paternal workplace control could independently affect maternal and paternal control on children.

In addition, while Colvin and Pauly (1983) do not specifically discuss the importance of the State on levels of control within the family-there is an implicit acknowledgement of the importance of the State and in the case of the United States, capitalism, on control in the family. Beyond workplace control, state control, for example in the form of welfare policy may be an important addition to power-control theory. Families in general, and single parent families in particular often find themselves negotiating more than just workplace control.

Finally, Tittle's (1995) control balance theory offers a second framework for consideration in future reformulations of power-control theory. To this stage, both relational and instrumental control have been specified as linearly related to risk preference, risk perception, and deviance-an alternative specification drawing from control balance theory (Tittle 1995; Piquero and Hickman 1999) would suggest that both a control deficit and overcontrol may lead to deviant behavior. While this is not the initial formulation of control in power-control theory, some researchers have implicitly suggested that high levels of control may lead to an increase in some types of deviance, specifically deviant adolescent role exits (Hagan 1990; Blackwell et al. 2002). This could have implications for further research on family structure and power-control theory because previous research has shown that changes in family formation (for example, moving from a single parent household to stepfamily household) can initially cause changes in child rearing practices (increases and decreases in control and supervision over children) as parents renegotiate their parenting roles (Papernow 1992). In addition, these changes often have different effects on sons and daughters in the household (Papernow 1992).

Such reformulations would not only have implications for how power-control theory might address family structure, but would also have implications for how one might examine gender differences in deviance. Disaggregating the levels of patriarchy in the household by looking at levels of workplace control, not as relational measures, but separate mechanisms may begin to tease out the importance of maternal and paternal controls separately. As this study has found, paternal controls do not always act in the same manner as maternal controls on gender and deviance. These findings also have implications for public policy.

Implications for Power-Control Theory and Public Policy
Finally, this study implicitly suggests practices to control deviant and delinquent behavior. First, Cullen et al. (1998) argue that there may be a new demand emerging for early intervention programs that target high-risk children and high-risk families. The findings of this study support this argument and go one step further. This study suggests that programs that support children and families, in general, may be useful in decreasing delinquency. According to this study, both single mother and single father households show lower levels of instrumental control than two parent households-in other words, a decreased ability to monitor children. In turn, high levels of instrumental control were linked to high levels of perceived risk by children and lower levels of deviance directly. This suggests that single parent families need more help monitoring their children. If levels of instrumental control (monitoring) could be increased in these households, levels of deviance should decrease. Formal policies that support all families through parenting classes to explain the importance of consistent monitoring or resources designed to increase the ease of monitoring (for example, policies that subsidize day care for those families who cannot afford it) would be consistent with this theory and these findings.

Second, researchers suggest that there has been a resurgence of propatriarchal sentiments in the U.S. as seen through such movements as the Promise Keepers and the National Fatherhood Initiative (Coltrane 2001; McCarthy et al. 1999) and that the implication of these movements must be evaluated with respect to such theories as power-control (McCarthy et al. 1999). While these movements profess a renewed commitment to family, this study suggests an important question-what does this renewed commitment look like?

For example, the Promise Keepers promotes fatherhood, but in a traditionally patriarchal way that places men at the head of the family and expects "gracious submissiveness" of wives (Coltrane 2001). While the teachings of this movement call for a renewed commitment by men to their families, this commitment is seen to flow "from God to men to women to children" (Coltrane 2001: 405), thereby suggesting that the daily childrearing is still the responsibility of women.

According to this study, patriarchal attitudes in the household might actually lead to less paternal control-perhaps because in patriarchal households the everyday parenting rests significantly with the mother. If this is the case, such movements that on the surface call for more male involvement in the family, should be analyzed for what this involvement really means.

Power-control theory would argue that programs and movements that offer real parenting support-in the form of: 1) helping to shape attitudes that call for hands on parenting from both mothers and fathers, 2) helping to shape attitudes that lead to the monitoring and control of both sons and daughters equally, and 3) offering resources to those, such as single parents, who may at times find daily monitoring of their children a challenge are important issues-and ones on which those wishing to reduce deviant behavior in both males and females should concentrate.

ENDNOTES
1 Unfortunately, these data do not allow the power-struggles that may exist between custodial and non-custodial parents or the custodial parent and the state and the effect these have on levels of patriarchy in the household to be teased out-these data do not have information on non-custodial parents or the state. However, given the previous research that has suggested the importance of extending our understanding of patriarchy beyond power differentials between custodial parents (see Morash & Chesney-Lind 1991 and Leiber & Wacker 1997), future research should also address the link between such outside forces as non-custodial parents on levels of patriarchy in the household. This direction may also illustrate further differences within single parent households and between single mother and single father households. back

2 Where possible, the same structural and attitudinal patriarchy measures as Grasmick et al. (1996) have been used (see Appendix).back

3 Though some have expressed a concern between patriarchy (regardless of measurement type) and higher income-such is not the case in these data. There is no correlation between family income and the global patriarchy measure in this study and there is a small but significant negative correlation of -.16 (p < .01) between family income and the attitudinal patriarchy measure. This is further evidence that while the sample may be over representative of the middle class - there is still a difference between class as measured by income and patriarchy in this sample.back

4 In much research, family structure has most often been treated as a static construct (Coontz 1997), by asking juveniles what family structure they grew up in. However, family structure is not static-a significant number of juveniles move through a variety of family forms or structures throughout their formative years. The best way to measure family structure would be to ask a series of questions that track all the stages of family formation a juvenile experiences, linking each of these stages to a series of dates (for an excellent example see the National Survey of Families and Households), then asking all relevant family questions for each of these stages. However, such an extended series of questions was beyond the scope of data collection for this project. Instead, we opted to choose a single point of time for juveniles to report their family structure, and then requested they answer the attitudinal patriarchy and maternal and paternal control questions with this family structure in mind. In this way, questions pertaining to parents could be linked to the family structure in question. According to researchers examining the differences between early onset delinquency and what is termed late onset delinquency, girls are rarely found to be life-course-persistent offenders (Moffit 1993) and are less likely to be found in the early onset category (Tibbetts & Piquero 1999). If researchers use retrospective questions for family structure focusing on a particular age, they most often use age 14 or 16 (Cherlin & Horiuchi 1980; Blackwell 2000). The authors chose 16. back

5 This is, again, a likely result of the sample type. It is probable that traditional households are overrepresented in a college sample, although the census reports that in the year 2000 69 percent of all family households were two parent, while 31 percent were one parent households (Fields & Casper 2001).back

6 Single father families have been notoriously underrepresented in research given their small numbers in the general population. While four percent is a small number, this study has not violated any statistical rules by including this dummy variable in the model (Allison 1999). In fact, the less the variance in a variable, the less its power to predict (Hardy 1993) which biases the likelihood of supporting the hypotheses. Because the sub-sample size of this group is so small, any non-findings should be viewed as preliminary. The authors chose to run analyses that included dummy variables instead of disaggregating the data by family structure type because the small number of single father families precluded conducting separate analyses.back

7 Analyses that further examine the extended family, such as grandparent living arrangements or non-custodial parenting are important-but beyond the scope of this paper.back

8 There is some concern that structural and attitudinal measures of patriarchy that are reported by the juvenile and not his/her parents cannot accurately portray levels of patriarchy in the family. Indeed, Davies and Kandel (1981) report that perceptions of parental educational aspirations for their children and the juvenile's perceptions of the same differ. While the best measure of patriarchy would probably be reported by the parents, Brewin et al. (1993) suggest that adults do accurately recall the details of their childhood. Moreover, it was beyond the scope of this study to collect data from both the respondent and his/her parents. A second concern with this measure is the retrospective design that requires the respondent to report on attitudes and behaviors in the respondent's past. Hagan (1989) suggests that when longitudinal data are not available, retrospective data are an acceptable alternative. This study used respondent reports of attitudinal patriarchy and a retrospective design that have previously been used with success by Grasmick et al. (1996).back

9 As Hagan et al. (1985) suggest, a theory such as power-control best explains non-serious forms of delinquency or deviance. Our measure, focusing on such behaviors as theft and drug use, are in keeping with the behaviors Hagan suggests power-control theory is most useful in explaining. We chose not to include one more measure of delinquency that Hagan and others have used before (drinking alcohol) because a small portion of our sample was over the age of 21, making this behavior non-deviant.back

10 The authors would have preferred to use multiple waves of interviews to ensure the casual ordering of variables in the model. However, the authors were required to administer anonymous questionnaires, negating the ability to track respondents at a later date. Although power-control theory provides clear hypotheses as to how familial relationships should impact deviance and those hypotheses guided the development of this model, it is certainly possible that deviance, itself, weakens family relationships. In the future, it is hoped that longitudinal data can further specify the relationship between family structure, patriarchy and deviance and test for recursive effects. Previous researchers (for example, Blackwell 2000) also using cross-sectional data, have called for longitudinal data that would ideally test a power-control model.back

11 Ordered logit was selected as the method of analysis for the outcome variable, deviance, because the variable is highly skewed to the left. Approximately 34.2 percent of respondents report no deviance, 28.4 percent one form of deviance, 23.7 percent two forms of deviance, 11.2 percent three forms of deviance and only 2.5 percent of the sample reported engaging in all four forms of deviance included. Ordered logit provides reasonable estimates for ordinal, dependent variables which do not have a normal distribution. We estimated this equation with OLS to compare standardized coefficients to the ordered logit models. We found no significant differences for any of the parameter estimates. Ordered logit does not report an r-squared or adjusted r-squared statistic, instead ordered logit reports a pseudo r-squared statistic. This is the statistic reported in Equation 8.back

REFERENCES
Aiken, Leona S. and Stephen G. West. 1991. Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: SAGE.

Allison, Paul D. 1999. Multiple Regression : A Primer. Thousand Oaks, California : Pine Forge Press.

Amato, Paul R. and Booth, A. 1991. "The Consequences of Divorce for Attitudes Toward Divorce and Gender Roles." Journal of Family Issues 12: 306-322.

Avakame, Edem F. 1997. "Modeling the Patriarchal Factor in Juvenile Delinquency: Is There Room For Peers, Church, and Television?" Criminal Justice and Behavior 24: 477-494.

Blackwell, Brenda Sims. 2000. "Perceived Sanction Threats, Gender, and Crime: A Test and Elaboration of Power-Control Theory." Criminology 38: 439-488.

Blackwell, Brenda Sims, Christine S. Sellers, and Sheila M. Schlaupitz. 2002. "A Power-Control Theory of Vulnerability to Crime and Adolescent Role Exits-Revisited." The Canadian Review of Sociology and Anthropology 39 (2): 199-218.

Blair, Johnny, Greta Menon, and Barbara Bickart. 1991. "Measurement Effects in Self vs. Proxy Responses to Survey Questions: An Information Processing Perspective." In Paul P. Biemer, Robert M. Groves, Lars E. Lyberg, Nancy A. Mathiowetz, and Seymore Sudman (eds.), Measurement Errors in Surveys. New York: Wiley.

Bloode, R., and D. Wolfe. 1960. Husbands and Wives. Glencoe, Illinois: Free Press.

Brewin, Chris R., Bernice Andrews, and Ian H. Gotlib. 1993. "Psychopathology and Early Experience: A Reappraisal of Retrospective Reports." Psychological Bulletin 113(1): 82-98.

Cherlin, Andrew J. 1992. Marriage, Divorce, Remarriage. Cambridge, Massachusetts: Harvard University Press.

Cherlin, Andrew J. and Shiro Horiuchi. 1980. "Retrospective Reports of Family Structure: A Methodological Assessment." Sociological Methods and Research 8: 454-469.

Coltrane, Scott. 2001. "Marketing the Marriage 'Solution': Misplaced Simplicity in the Politics of Fatherhood." Sociological Perspectives 44 (4): 387-418.

Colvin, Mark and John Pauly. 1983."A Critique of Criminology: Toward an Integrated Structural-Marxist Theory of Delinquency Production." American Journal of Sociology 89: 513-551.

Coontz, Stephanie. 1997. The Way We Really Are: Coming to Terms with America's Changing Families. New York: Basic Books.

Cullen, Francis T., John Paul Wright, Shayna Brown, Melissa M. Moon, Michael B. Blankenship, Brandon K. Applegate. 1998. "Public Support for Early Intervention Programs: Implications for a Progressive Policy Agenda." Crime and Delinquency 44 (2): 187-204.

Davies, Mark and Denise B. Kandel. 1981. "Parental and Peer Influences on Adolescents' Educational Plans: Some Further Evidence." American Journal of Sociology 87 (2): 363-387.

Demo, David H. and Alan C. Acock. 1988. "The Impact of Divorce on Children." Journal of Marriage and the Family 50: 619-648.

Dowd, Nancy E. 1997. In defense of single-parent families. New York: New York University Press.

Fields, Jason and Lynne M. Casper. 2001. "American's Families and Living Arrangements: Population Characteristics." Current Population Reports, P20-537, U.S. Census Bureau, Washington D.C.

Finlay, Barbara, Charles E. Starnes, and Fausto B. Alvarez. 1985. "Recent Changes in Sex-Role Ideology among Divorced Men and Women: Some Possible Causes and Implications." Sex Roles 12: 637-654.

Ganong, Lawrence and Marilyn Coleman. 1984. "Effects of Remarriage on Children: A Review of the Empirical Literature." Family Relations 33: 389-406.

Garasky, Steven, and Daniel R. Meyer. 1996. "Reconsidering the increase in father-only families." Demography 33: 385-394.

Goode, William. 1982. "Why Men Resist" in Rethinking the Family: Some Feminist Questions, edited by Barrie Thorne and Marilyn Yalom. New York: Longman.

Gottfredson, Michael and Travis Hirschi. 1990. A General Theory of Crime. Stanford, California: Stanford University Press.

Grasmick, Harold G., John Hagan, Brenda Sims Blackwell, and Bruce J. Arneklev. 1996. "Risk Preferences and Patriarchy: Extending Power-Control Theory." Social Forces 75 (1): 177-199.

Hagan, John. 1989. Structural Criminology. Rutgers University Press.

--------. 1990. "The Structuration of Gender and Deviance: A Power-Control Theory of Vulnerability to Crime and the Search for Deviant Role Exits." The Canadian Review of Sociology and Anthropology 27 (2): 137-156.

-------- and Fiona Kay. 1990. "Gender and Delinquency in White-Collar Families: A Power-Control Perspective." Crime and Delinquency 36: 391-407.

-------- and Blair Wheaton. 1993. "The Search for Adolescent Role Exits and the Transition to Adulthood." Social Forces 71(4): 955-980.

--------, A. R. Gillis, and John Simpson. 1985. "The Class Structure of Gender and Delinquency: Toward a Power-Control Theory of Common Delinquent Behavior." American Journal of Sociology 90: 1151-1178.

--------, A. R. Gillis, and John Simpson. 1993. "The Power of Control in Sociological Theories of Delinquency." In Freda Adler and William Laufer (eds.), New Directions in Criminological Theory. New Brunswick, N.J.: Transaction.

--------, A. R. Gillis, and John Simpson. 1990. "Clarifying and Extending Power-Control Theory." American Journal of Sociology 95 (4): 1024-1037.

--------, John Simpson, and A. R. Gillis. 1987. "Class in the Household: A Power-Control Theory of Gender and Delinquency." American Journal of Sociology 92 (4): 788-816.

Hardy, Melissa A. 1993. "Regression with Dummy Variables." Sage Series: Quantitative Applications in the Social Sciences Volume 93.

Heimer, Karen. 1996. "Gender, Interaction, and Delinquency: Testing a Theory of Differential Social Control." Social Psychology Quarterly 59: 39-61.

Hill, G. D. and M. P. Atkinson. 1988. "Gender, Familial Control, and Delinquency." Criminology 26: 127-147.

Hindelang, Michael J., Travis Hirschi, and Joseph G. Weis. 1981. Measuring Delinquency. Beverly Hills, California: Sage.

Hirschi, Travis. 1969. Causes of Delinquency. University of California Press.

Jensen, Gary. 1993. "Power-Control vs. Social Control Theories of Common Delinquency: A Comparative Analysis." Pp. 363-380 in New Directions in Criminological Theory, edited by Freda Adler and William Laufer. Transaction Publishers.

Jensen, Gary F. and Kevin Thompson. 1990. "What's Class Got to Do with It? A Further Examination of Power-Control Theory." American Journal of Sociology 95 (4): 1009-1023.

Lieber, Michael J. and Mary Ellen Ellyson Wacker. 1997. "A Theoretical And Empirical Assessment of Power-Control Theory and Single mother Families." Youth and Society 28 (3): 317-350.

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks: Sage Publications.

McCarthy, Bill and John Hagan. 1987. "Gender, Delinquency, and the Great Depression: A Test of Power-Control Theory." Canadian Review of Sociology and Anthropology 24: 153-177.

McCarthy, Bill, John Hagan and Todd S. Woodward. 1999. "In the Company of Women: Structure and Agency in a Revised Power-Control Theory of Gender and Delinquency." Criminology 37: 761-788.

McKelvey, R. D. and W. Zavoina. 1975. "A Statistical Model for the Analysis of Ordinal Level Dependent Variables." Journal of Mathematical Sociology 4: 103-120.

Moffit, Terrie E. 1993. "Adolescent-Limited and Life-Course Persistent Antisocial Behavior: A Developmental Taxonomy." Psychological Review 100: 674-701.

Morash, Merry and Meda Chesney-Lind. 1991. "A Reformulation and Partial Test of the Power-Control Theory of Delinquency." Justice Quarterly 8 (3): 347-377.

Papernow, Patricia L. 1992. Becoming a Stepfamily. San Francisco: Joey-Bass Publishers.

Parker, Gordon. 1989. "The Parental Bonding Instrument: Psychometric Properties Reviewed." Psychiatric Developments 4: 317-335.

Pedhauzer, Elazur. 1982. Multiple Regression in Behavioral Research 2nd Ed. Orlando, FL: Harcourt, Brace College Press.

Piquero, Alex R. and Matthew W. Hickman. 1999. "An Empirical Test of Tittle's Control Balance Theory." Criminology 37 (2): 319- 340.

Richmond-Abbott, Marie. 1984. "Sex-Role Attitudes of Mothers and Children in Divorced Single-Parent Families." Journal of Divorce 8 (1): 61-81.

Robins, Lee N., S. Schoenberg, S. Holms, K. Ratcliff, A. Benham, and J. Works. 1985. "Early Home Environment and Retrospective Recall." American Journal of Orthopsychiatry 55: 27-41.

Sacco, Vince. 1990. "Gender, Fear, and Victimization: A Preliminary Application of Power-Control Theory." Sociological Spectrum 10: 485-506.

Singer, Simon I., and Murray Levine. 1988. "Power-Control Theory, Gender, and Delinquency: A Partial Replication with Additional Evidence on the Effects of Peers." Criminology 26 (4): 627-647.

Slavkin, Michael L. 2001. "Gender Schematization in Adolescents: Differences Based on Rearing in Single-Parent and Intact Families." Journal of Divorce and Remarriage 34: 137-149.

Slavkin, Michael and Anne Dopkins Stright. 2000. "Gender Role Differences in College Students from One- and Two-Parent Families." Sex Roles 42: 23-37.

Thomson, E., Sarah McClanahan, and R. Curtin. 1992. "Family Structure, Gender, and Parental Socialization." Journal of Marriage and the Family 54: 368-378.

Tibbetts, Stephen G. and Alex R. Piquero. 1999. "The Influence of Gender, Low Birth Weight, and Disadvantaged Environment in Predicting Early Onset of Offending: A Test of Moffitt's Interactional Hypothesis." Criminology 37(4): 843-878.

Tittle, Charles. 1995. Control-Balance: Toward a General Theory of Deviance. Boulder, Colorado: Westview.

Wright, David W. and Robert Young. 1998. "The Effects of Family Structure and Maternal Employment on the Development of Gender-Related Attitudes among Men and Women." Journal of Family Issues 19: 300-314.

ABOUT THE AUTHORS
Kristin A. Bates is an Assistant Professor in the Department of Sociology at California State University, San Marcos. Her research interests include family and delinquency, and gender and racial discrimination within the criminal justice system.

Christopher D. Bader is an Assistant Professor in the Department of Sociology and Anthropology at Baylor University. His research interests include new theoretical frameworks of crime and criminology. Other research interests include control theories of delinquency and religion. Christopher D. Bader, Department of Sociology and Anthropology, Baylor University, Waco, Texas, 76798.

F. Carson Mencken is an Associate Professor in the Department of Sociology and Anthropology at Baylor University. His research interests include spatial patterns of economic growth and development, crime and delinquency, and quantitative methods. F. Carson Mencken, Department of Sociology and Anthropology, Baylor University, Waco, Texas 76798. back

Contact Information
Direct correspondence to Kristin A. Bates, Department of Sociology, CSUSM, San Marcos, California, 92096-0001. kbates@csusm.edu.

APPENDIX: Question Wording
The following are the complete question wordings for items used to create the measures of attitudinal patriarchy, instrumental control, relational control, risk preference, and perceived risk. These questions have been taken from survey questions initially used by Grasmick et al. (1996).

Attitudinal Patriarchy
Respondents were asked the following questions about each parent. Response categories were "Strongly Agree," "Agree," "Disagree," and "Strongly Disagree." Items were reverse-coded if necessary, such that higher scores represented more patriarchal attitudes.

• How would your [mother/father] have responded to the following: It is okay for a mother to work full-time when their youngest child is under age five.
• How would your [mother/father] have responded to the following: Preschool children are likely to suffer if their mother is employed.
• How would your [mother/father] have responded to the following: It is difficult for young children when their mother is employed full-time.
• How would your [mother/father] have responded to the following: Parents should encourage just as much independence in their daughters as their sons.
• How would your [mother/father] have responded to the following: Mothers should encourage their daughters to seek a career just as much as their sons.
• How would your [mother/father] have responded to the following: It is much better for everyone if the man earns the main living and the woman takes care of the home and family.
• How would your [mother/father] have responded to the following: If a husband and a wife both work full-time, they should share household tasks equally.
• How would your [mother/father] have responded to the following: If a woman is offered a promotion, her husband should be willing to move for the sake of her career.
• How would your [mother/father] have responded to the following: Men are by nature better leaders for the family than are women.

Instrumental Control
Respondents were asked the following questions about each parent. Response categories were "Strongly Agree," "Agree," "Disagree," and "Strongly Disagree." Items were reverse-coded, such that higher scores represent greater instrumental control.

• My mother/father generally knew where I was when I was away from home.
• My mother/father generally knew whom I was with when I was away from home.

Relational Control
Respondents were asked the following questions about each parent.

• Do you talk with your [mother/father] about your thoughts and feelings?
  Response categories: "Never," "Sometimes," "Usually," "Always."
• Would you like to be the kind of person your mother is?
  Response categories: "Not at all," "In a few ways," "In some ways," "In most ways," "In every way"

Perceived Risk
Response categories were "Strongly Agree," "Agree," "Disagree," and "Strongly Disagree." Items were reverse-coded, such that higher scores represent greater perceived risk.

• I could break into a place and not get caught.
• I could steal from a store and not get caught.
• I could write graffiti and not get caught.

Risk Preference
Response categories were "Strongly Agree," "Agree," "Disagree," and "Strongly Disagree." Items were reverse-coded, such that higher scores represent greater risk preference.

• I like to test myself every now and then by doing something a little risky
• I sometimes find it exciting to do things for which I might get in trouble
• Excitement and adventure are more important to me than security