Gender Double Standards in Parenting Attitudes

Abstract

This paper investigates the double standard in attitudes toward courtship and family formation behaviors of sons and daughters. We argue there are strong theoretical reasons to expect that the magnitude of this double standard varies across substantive domains, as well as amongst parents and non-parents. We also argue key methodological limitations of previous studies likely produce an under-estimate of the gender double standard. We provide empirical estimates of the gender double standard that overcome these limitations, including a random assignment experiment explicitly designed to control the effects of social desirability. These estimates demonstrate variability in the double standard across domains and reveal key factors contributing to the magnitude of the double standards in parenting attitudes held by individuals.

Gender differences in social behaviors are among the most highly studied topics in the social sciences. Research emphasizes changes over time, causes, and consequences related to the gender segregation in social roles. Over the last half of the 20^th century social research documented wide spread changes in attitudes about the appropriate social roles for men and women, uniformly toward more gender egalitarian attitudes (Brewster and Padavic 2000; Brooks and Bolzendhal 2004; Bumpass 1990; Cherlin and Walters 1981; Mason and Lu 1988; Thornton 1989; Thornton and Young-DeMarco 2001; Veroff, Douvan and Kulka 1981). However, recent research documents the persistence of substantial gender differences in key family formation behaviors over the same period. This includes gender differences in sexual, cohabiting, marital, and childbearing behaviors (Barber 2001a; Laumann et al. 1994; Marini 1978; Michael and Tuma 1985; Rindfuss, Morgan and Swicegood 1988; Smock 2000; Thornton, Axinn and Xie 2007; Waite and Goldscheider 1991). In this paper we investigate a set of attitudes that may contribute to these substantial gender differences in family behaviors.

We argue that ideational factors may be an important reason for gender differences in family behaviors in spite of the well known trend toward more egalitarian gender role attitudes. Other attitudes may be at work. For example, a substantial body of research demonstrates that courtship and family formation attitudes and beliefs are strongly correlated with subsequent cohabiting, marital, and childbearing behaviors in the United States and Europe (Axinn and Thornton 1992, 1993; Barber 2000, 2001b; Barber and Axinn 2004; Clarkberg, Stolzenberg and Waite 1995; Lesthaeghe 1983, 2002; Lesthaeghe and Surkyn 1988; Plotnick 1992). This research points toward parents’ attitudes as a particularly powerful force in shaping their children’s behaviors with respect to sex, cohabitation, marriage and childbearing (Axinn and Thornton 1993; Barber 2000). Moreover, this same body of research shows that men and women respond in similar ways to their parents’ attitudes, so there are not substantial gender differences in correlation between parental attitudes and children’s subsequent behavior (Axinn and Thornton 2003; Barber 2000). However, we argue that there may be substantial gender based differences in the parental or prospective parental attitudes about courtship and family formation known to predict children’s subsequent behavior.

Unfortunately, such gender based differences in parental or prospective attitudes towards their children’s behaviors have received very little scientific attention. We investigate gender differences in parental preferences for children’s courtship and family formation behaviors, or what we refer to as the gender double standard in parenting attitudes. We begin by formulating a conceptual framework for the study of this double standard that identifies the breadth of the double standard in parenting values among parents and non-parents. Next, we explain why the magnitude of this double standard is likely to vary across topics, even within the domain of courtship and family formation. Then we identify the key methodological issues limiting our ability to detect these double standards. Finally, we use measures from a unique random-assignment experiment to separate social desirability effects on responses from the true parenting gender double standard. This approach allows us to document the magnitude of both the social desirability effect and the underlying gender double standard in parenting values. These unusual measures also provide the means to examine differences in these parenting double standards across substantive domains. Together these advances create substantial new insight into the processes producing gender double standards in parenting values.

Theoretical and Methodological Background

At the foundation of our conceptual framework is a great deal of previous research on the origins of gender differences in attitudes and behavior. Gender differences in social life have been a focus of sociological and psychological theory throughout the history of these disciplines. Of key interest is whether and to what degree the social differences observed between men and women (e.g., education, occupation, marriage) are influenced by the way the sexes are differentially treated in various social environments and across the life course. For example, many scholars contend that the unequal treatment of men and women can be traced all the way back to different environments at birth, suggesting that the gender differences we observe later in life are a direct result of the distinct behavior, attitudes, and expectations we maintain toward infant boys and girls (Grieshaber 1998; Mondschein, Adolph and Tamis-LeMonda 2000; Pomerleau et al. 1990; Seavey, Katz and Zalk 1975; Sidorowicz and Lunney 1980). Throughout childhood, others observe that historically defined masculine behaviors tend to be enforced or permitted in boys and more feminine behaviors are encouraged in girls (Antill 1987; Crowley et al. 2001; Fagot 1974; Hoffman 1977; Martin 1998; Price-Bonham and Skeen 1982). Even when there is no evidence of gender enforced behavior or attitudes, many studies nevertheless reveal that mothers and fathers retain some distinct gender role attitudes or parenting styles that may vary in magnitude depending on the sex of the child (Conrade and Ho 2001; Grieshaber 1998; McGovern 1990; Ricks 1985; Rossi 1984; Rothbart and Maccoby 1966) and that parents frequently differentiate between their male and female children leading to unequal treatment (Antill 1987; Rossi 1984). Building on these studies, we construct a framework below that defines a gender double standard in parenting values, considers variations in those double standards across subjects, and addresses the fundamental methodological issues involved in studying these double standards.

Defining the Gender Double Standard

We focus on the gender double standard in attitudes toward courtship and family formation behavior. A double standard “implies that two things which are the same are measured by different standards” (Eichler 1980:15). Thus a gender double standard suggests that we evaluate the same behavior of men and women differently; what is acceptable or appropriate for one may not be equally so for the other. However, the gender double standard, applicable towards any type of behavior, has largely been discussed in previous research within the context of the acceptability of premarital sex (Modell 1989; Reiss 1960, 1964, 1967). The original operationalization offered by Reiss (1960, 1964, 1967) describes the double standard as the different standards of sexual permissiveness for women and men.

Since his work, the evidence regarding the existence of a gender double standard has been inconclusive. Some findings reveal that women who engage in premarital sex are more negatively evaluated than are men for the same behavior (Crawford and Popp 2003; Galper and Luck 1980; Harrison, Bennett and Globetti 1969; Milhausen and Herold 1999; Oliver and Sedikides 1992; Reiss 1967; Sheeran et al. 1996; Sprecher and Hatfield 1996; Treboux and Busch-Rossnagel 1990). Other scholars, however, contend that there is little to no evidence for the existence of a gender double standard (DeLameter and MacCorquodale 1979; Gentry 1998; Jacoby and Williams 1985; King, Balswick and Robinson 1977; Mark and Miller 1986; O’Sullivan 1995; Peplau, Rubin and Hill 1977; Sprecher 1989). And some have either been unable to find evidence for or against its existence (Istvan and Griffitt 1980) or believe that the nature of this double standard has evolved so that detecting it has become much more difficult (Sprecher, McKinney and Orbuch 1987). Interestingly, one study by Milhausen and Herold (1999) found that regardless of whether a gender double standard actually remains, the women in their study overwhelming believed that a double standard for sexual behavior (where it is more acceptable for men to have more sexual partners than it is for women) does exist in society.

We believe a number of factors contribute to these inconsistent findings regarding the gender double standard. They include differences in the subject matter of the gender double standard, generality of the gender comparison, the growing social desirability of presenting no double standard, and limitations of, or differences in the sample designs employed in previous research.

Variations in Gender Double Standards across Topics

We argue there are strong theoretical reasons to expect the gender double standard to vary across subject matter. We focus our investigation on courtship and family formation processes, but even within this domain we expect the gender double standard to vary across topic. We differentiate dating, sex, and premarital cohabitation from marital and childbearing behavior to investigate differences within the courtship and family formation domain.

Many factors point toward especially strong gender double standards in parental attitudes about their children’s courtship. Historically the United States has been characterized by stronger prohibitions on women’s participation in premarital sex (Modell 1989; Tolman 1991), so we expect tolerance of early dating, early sex, and premarital cohabitation will be significantly higher for sons than for daughters (Laumann et al. 1994; Moore and Stief 1991; Thornton and Young-DeMarco 2001; Thornton et al. 2007). In part, this difference may be purely the product of gender based ideologies of protecting daughters’ sexual purity (Ferree 1990). For example, Treboux and Busch-Rossnagel (1990) find that the message sons and daughters receive from parental discussions differ such that parents appear accepting of sexuality in their sons, “The message conveyed to daughters may be ‘Don’t–and if you do, we don’t want to know about it,” (Treboux and Busch-Rossnagel 1990:185). Research on teenage sexuality indicates parents communicate more factual information and moral discussion with their daughters about sex and dating than with sons (Nolin and Petersen 1992; Wood et al. 2002). Some scholars even argue that teenage girls are taught to repress or silence their sexual feelings or desire, a finding that is consistent with the popular view of women not engaging in sexual behavior outside of marriage (Tolman 1994).

Of course the difference in attitudes about sons and daughters’ sexuality may also be based on parental concerns regarding the potential consequences of a daughter’s pregnancy (Schalet 2000). The evidence of long term consequences for young women of pregnancy and birth before marriage or at young ages has been a matter of substantial debate (Geronimus and Korenman 1992; Hoffman 1998; Klepinger, Lundberg and Plotnick 1995). But for the parents of young women the short term consequences may be clearer. Young unwed mothers are highly likely to live with their parents and parents are quite likely to make financial contributions to the care of such grandchildren (Aquilino 1996; Sigle-Rushton and McLanahan 2002). The potential burden of a daughter becoming pregnant may be more likely to fall on the parents than the burden of a son causing a pregnancy. As a result there may be cost/benefit reasons for this gender double standard in addition to potential ideological reasons.

Parental double standards regarding their children’s courtship may also vary by the nature of the courtship. Although we do not explicitly evaluate the gender double standard based on the seriousness or stage of a relationship, previous research has shown that the gender double standard is stronger the more casual the relationship (Reiss 1967; Roche 1986; Sprecher and Hatfield 1996). Therefore we expect that the double standard will be more pronounced for sex and dating than for premarital cohabitation.

The historical context of ideas about men’s and women’s behavior in the United States is likely to shape family formation double standards in the opposite direction. The United States is characterized by a longstanding acceptance of earlier marriage for women, so we expect Americans to hold lower ideal ages for women to marry and to place less importance on women working before marriage than men (Thornton and Young-DeMarco 2001). So even though double standards in courtship are in the direction of later courtship for women, we expect the double standards in marriage to be toward early marriage for women. This contrast is consistent with both ideological differences in the concern of sexual purity of sons and daughters and potential gender based differences in the parental burden associated with a child becoming pregnant or causing a pregnancy. Later courtship and early marriage mean daughters are less likely to have sex or become pregnant while unmarried.

Finally, separate from the timing of marriage, we expect little gender based double standard in respondents’ ideas about their children eventually marrying and having children. In part this is because marriage and childrearing continue to be strongly endorsed among both men and women (Moore and Stief 1991; Rossi 1984; Thornton and Young-DeMarco 2001). Likewise, remaining childless and unmarried has been unpopular for both men and women (Blake 1979; Polit 1978; Thornton 1989; Veroff et al. 1981). Thus we have little basis to expect to find gender differences in this domain.

Methodological Issues in Studying the Gender Double Standard

Previous research documenting the gender double standard may not reveal its full extent for three key methodological reasons. First, research in this area has often been limited to the evaluation of a general, unspecified “other” – a fictional woman or man who is occasionally defined as a college student or classmate (Jacoby and Williams 1985; Mark and Miller 1986; O’Sullivan 1995; Reiss 1967; Sheeran et al. 1996; Sprecher 1989; Sprecher et al. 1987; Sprecher et al. 1988), and less often limited to a hypothetical friend, future date, or future spouse (Milhausen and Herold 1999; Oliver and Sedikides 1992). We argue the double standard is most clearly revealed when asking about someone with strong social ties, even if the person is hypothetical. We believe this is especially true when those ties are one’s own children, whether those children exist already or may be present sometime in the future. Although few studies indicate the magnitude of a sexual double standard among parents, it is clear from previous research on parenting that parents do indeed hold a gender double standard for their children in other domains such as gender role attitudes and the appropriateness of non-sexual behaviors (Crowley et al. 2001; Pomerleau et al. 1990) and may even convey different messages to daughters and sons about certain behavior such as sex (Treboux and Busch-Rossnagel 1990).

Second, many studies are based on samples with limited representation, quite often focusing on college students (Galper and Luck 1980; Hendrick and Hendrick 1987; Hong 1983; Jacoby and Williams 1985; Kinnaird and Gerrard 1986; Maranell et al. 1970; Mark and Miller 1986; Sprecher and Hatfield 1996; Sprecher et al. 1987). College students are known to hold attitudes that differ systematically from those in the general adult population (see Sears 1986 for a discussion). Both the processes of selection into institutions of higher education and the social environments of these institutions are likely to produce smaller gender double standards (Alwin et al. 1991). We argue that the adult population of the United States is more likely to possess a strong gender double standard than would college students.

Third, all research involving human subjects may be affected whenever the conditions or topic of the study can potentially affect the respondent to want to present him or herself more positively to the researcher. Social desirability, or “the tendency of people to deny socially undesirable traits or qualities and to admit socially desirable ones” (Phillips and Clancy 1972: 923), is a well-known threat to the measurement of behavior and attitudes in many domains (Belli et al. 1999; Bishop, Tuchfarber and Oldendick 1986; Crowne and Marlowe 1964; Phillips and Clancy 1972; Press and Townsley 1998; Presser 1990; Rossiter and Robertson 1975; Theriault and Holmberg 1998). This area of research reveals that social desirability tends to be a problem for accurate measurement regardless of how the information is collected, with self-reports or mail-in surveys probably yielding the lowest amount of this kind of bias (Press and Townsley 1998) and face-to-face interviews yielding more (Aquilino and Lo Sciuto 1990; Finkel, Guterbock and Borg 1991; Krysan et al. 1994). In addition, the level of social desirability bias present is also affected when what is being measured is of a slightly controversial or sensitive nature such as racial attitudes or premarital sexual behavior (Aquilino and Lo Sciuto 1990; Krysan 1998; MacCorquodale and DeLamater 1979). As gender role attitudes have become more egalitarian in the United States there is growing recognition that it is socially desirable to present no gender double standard when asked about courtship and family formation choices (Thornton and Young-DeMarco 2001). Without explicit steps designed to address the issue of social desirability in responses, research is likely to under-estimate the magnitude of the gender double standard.

To address these three obstacles to the measurement of the gender double standard we use three specific strategies. First, we focus on gender double standards for the behavior of one’s own children (both actual and hypothetical) in order to measure the double standard for a group to whom the individual has extremely close social ties. Second, we measure the double standard in an adult general population sample. Third, we use a random assignment experiment to measure the level of social desirability and to account for that social desirability in our estimates of the magnitude of the gender double standard.

Data and Methods

Data for our analysis comes from a systematic, population sample of white 31-year olds. These respondents were chosen from a randomized sample of 1961 birth records from the Detroit Metropolitan area, and interviewed in 1993 when they were all approximately 31 and a half years of age. This sample design has the advantage of eliminating variance in age, place of birth, and race to assist in focusing on variance in other key factors such as gender, parenthood, and education. Of course inference cannot be made to groups not represented in the sample, including those of other ages, other places of birth, or other races. However, compared to previous research on these topics using college student based samples or other convenience samples, this birth record based sample has the advantage that the only other selection criteria is the occurrence of the individual’s birth itself.

Although the sample respondents were all born in one metropolitan area, by the time they reached adulthood most were living elsewhere, scattered throughout Michigan as well as the rest of the United States. Previous analyses concerning background characteristics, attitudes, and behaviors of these young adults have yielded results very similar to those derived from national samples (Thornton and Axinn 1996). Response rates for this group are excellent. Accounting for both initial selection in 1961 and tracking from 1962 to 1993, the response rate for this sample is approximately 83%, calculated by the number of 1993 interviews divided by the number of possible interviews in 1993 (defined as 1962 respondents minus the number who had died or were permanently ill). This yields 904 respondents.

These 904 respondents are, of course, diverse with respect to many of the key factors of interest. Approximately 51% are female, 59% have experienced parenthood, and educational attainment ranges from 8 years to 17 or more years. Of those who are parents, the maximum number of children is 6, the mean age of the oldest child is nearly 6 years old, and approximately 67 percent have at least one daughter. Table 2, which is explained in more detail below, presents descriptive statistics for all the measures included in our analyses separately for males and females and pooled together. Note that because the children of respondents who are currently parents are quite young, responses to questions about dating, sex, cohabitation, marriage, and childbearing reflect feelings about possible future behaviors of children among both parents and non-parents in this sample.

Table 2.

Means, Standard Deviations, and Ranges For Measures used in Analyses

	All Respondents				Males		Females
	Mean	Std. Dev	Min	Max	Mean	Std. Dev	Mean	Std. Dev
Attitudes Towards Courtship
Dating Young OK	3.73	1.66	1	8	3.95	1.83	3.51	1.44
Sex Young OK	3.95	1.99	1	8	4.25	2.04	3.65	1.90
Cohabitation OK	2.83	1.13	1	4	2.88	1.11	2.79	1.15
Timing of Marriage
Ideal Age at Marriage (Years)	26.07	2.86	18	35	25.53	2.88	26.58	2.75
Work Before Marriage Important	3.75	0.57	1	4	3.69	0.63	3.80	0.50
Marriage
OK if No Marriage	2.77	1.04	1	4	2.91	1.01	2.64	1.06
OK if No Children	2.67	1.03	1	4	2.78	0.99	2.57	1.06
Child’s Gender
If Son Record	.50	.50	0	1	.50	.50	.50	.50
CONTROLS
Social Desirability
If Daughter Asked 1st	.51	.50	0	1	.51	.50	.50	.50
Respondent Gender
If Respondent is Male	.49	.50	0	1
Respondent Family Formation Experience
Currently Married	.66	.47	0	1	.63	.48	.68	.47
Previously Married	.10	.30	0	1	.08	.27	.13	.34
Ever Cohabited	.51	.50	0	1	.50	.50	.51	.50
Ever a Parent	.59	.49	0	1	.53	.50	.64	.48
Had Sex by age 18	.60	.49	0	1	.63	.48	.58	.49
Has Daughters (Parents Only)	.67	.47	0	1	.63	.48	.70	.46
Age of Oldest Child (Parents Only)	5.98	3.83	0	17	5.63	3.71	6.27	3.90
# of Children Ever Born (Parents Only)	1.90	.86	1	6	1.84	.87	1.94	.85
Respondent Education, Work, Income
Education (years)	13.99	2.10	8	17	13.92	2.22	14.06	1.97
Work (% of last 6 months full-time)	77.73	36.49	0	100	92.73	20.62	63.26	42.19
Income (log 1993 family income)	10.53	1.44	0	13.38	10.55	1.24	10.51	1.60
Respondent Religious Background
Fundamentalist Protestant	.11	.31	0	1	.09	.28	.13	.33
Catholic	.54	.50	0	1	.57	.50	.52	.50
Jewish, Other, None	.12	.32	0	1	.14	.35	.09	.28
Religious Service Attendance	3.21	1.50	1	6	3.04	1.50	3.37	1.49

The 1993 interview included a set of child-rearing measures specifically designed to tap into the respondent’s attitudinal differences based on the gender of a child. Table 1 describes the actual wording and coding schemes associated with each measure. Each respondent, regardless of parental status, was asked to indicate how much he/she would be bothered by his/her child dating or having sexual intercourse at various ages, living with an opposite sex partner outside of marriage, not ever marrying, and not ever having children. The least restrictive views about a behavior were given the highest score possible (4 or 8), while the most restrictive attitudes were given a score of 1 (the lowest score possible). Respondents were also asked the ideal age for their daughter/son to marry. The actual age supplied by the respondent was used for the response code. Additionally, respondents were asked to rate the importance of working full time before marriage. This time the answer reflecting later marriage (extremely important) was scored highest, and the answer consistent with early marriage (not at all important) was scored lowest.

Table 1.

Dependent Variable Measures for Respondent Attitudes Toward Daughter and Son Behaviors

Attitudes Towards Courtship	Question Wording and Coding
Dating Young OKa	If a daughter/son of yours starts dating at 16/12, would that bother you a great deal, some, a little, or not at all? 1=A great deal at 16, 2=Some at 16, 3=A little at 16, 4=Not at all at 16, 5=A great deal at 12, 6=some at 12, 7=A little at 12, 8=Not at all at 12
Sex Young OKb	If a daughter/son of yours has sexual intercourse at age 20/16, would that bother you a great deal, some, a little, or not at all? 1=A great deal at 20, 2=Some at 20, 3=A little at 20, 4=Not at all at 20, 5=A great deal at 16, 6=some at 16, 7=A little at 16, 8=Not at all at 16
Cohabitation OK	Suppose that after she/he has grown up she/he decides to live with a man/woman in an intimate relationship without being married to her/him? Would that bother you a great deal, some, a little, or not at all? 1=A great deal, 2=Some, 3=A little, 4=Not at all
Timing of Marriage
Ideal Age (Years)	If a daughter/son of yours does get married, and if it were just up to you, what do you think would be the ideal age for her/him to get married? Code actual age
Work Before Marriage Important	What about work – how important do you think it would be for a daughter/son of yours to work full time for a year or two before she/he gets married – would you say very important, somewhat important, not very important, or not at all important? 1=Not at all important, 2=Not very important, 3=Somewhat important, 4=Very important
Marriage and Children
OK if No Marriage	Suppose that things turn out so that a daughter/son of yours does not marry, would that bother you a great deal, some, a little, or not at all? 1=A great deal, 2=Some, 3=A little, 4=Not at all
OK if No Children	Suppose that things turn out so that she/he does not have any children, would that bother you a great deal, some, a little, or not at all? 1=A great deal, 2=Some, 3=A little, 4=Not at all

^aRespondents were initially asked how much it would bother them if their daughter/son began dating at age 14. Those answering “a great deal” or “some” were asked an identical set of questions about dating at age 16. Those answering “a little” or “not at all” were asked an identical set of questions about dating at age 12. The dependent variable was then created by recoding the second response.

^bRespondents were initially asked how much it would bother them if their daughter/son had sexual intercourse at age 18. Those answering “a great deal” or “some” were asked an identical set of questions about intercourse at age 20. Those answering “a little” or “not at all” were asked an identical set of questions about intercourse at age 16. The dependent variable was then created by recoding the second response.

Note that questions about dating actually began with a question about dating at age 14 and questions about sex began with a question about sex at age 18. Those who replied that dating at age 14 would bother them a great deal or some were asked about dating at age 16, those who replied that dating at age 14 would bother them a little or not at all were asked about dating at age 12. Figure 1 presents these questions and response categories as they were in the questionnaire to demonstrate how the measures are linked and then coded as a single item. Similarly, those who replied that sex at age 18 would bother them a great deal or some were asked about sex at age 20, those who replied that sex at age 18 would bother them a little or not at all were asked about sex at age 16. Thus the items for age of dating and first sex in Table 1 represent the combination of respondents’ initial answers about a central age and respondents’ secondary answers to questions designed to create a wider distribution as shown in Figure 1.

Figure 1.

To check the sensitivity of our results, in addition to the analyses of the items presented in Table 1 and discussed throughout the text, we also analyzed the responses to the initial items on central ages (age 14 for dating and age 18 for sex). The correlations between responses to these initial items and the full scales are actually quite high (Pearson’s r = .81 for dating and .89 for sex), so it is no surprise analyses of these initial items yielded results substantively identical to those presented in the text. Of course these metrics for dating at 14 and sex by age 18 are different than those presented in Table 1. Responses to these initial questions are coded to range from 1–4, whereas the full range used in Table 1 varies from 1–8. This greater range produces some minor differences in coefficient sizes and specific estimates. However, for all the analyses presented in Tables 3–5 use of the initial item for central ages produces exactly the same conclusions about direction, standardized magnitude, and statistical significance as the full range measures actually presented in Tables 3–5.

Table 3.

Mean Differences toward Son and Daughter Family Formation Behavior Attitudes For All Respondents, Non-Parents, and Parents

	All Respondents N=904			Non-Parents N=366			Parents N=538
	Son	Dau	Diff	Son	Dau	Diff	Son	Dau	Diff
Courtship
Dating Young OK	4.02	3.43	0.59***	4.32	3.74	.58***	3.81	3.22	.59***
Sex Young OK	4.24	3.65	0.59***	4.61	4.04	.58***	4.00	3.34	.61***
Cohabitation OK	2.96	2.71	0.25***	3.19	3.00	.19***	2.80	2.51	.28***
Timing of Marriage
Ideal Age at Marriage (Years)	26.49	25.64	0.85***	27.39	26.36	1.02***	25.88	25.15	.74***
Work before Marriage Important	3.82	3.67	0.15***	3.82	3.67	.14***	3.83	3.67	.16***
Marriage and Children
OK if No Marriage	2.80	2.74	0.06**	3.04	2.88	.16***	2.64	2.64	.00
OK if No Children	2.69	2.66	0.02	3.01	3.01	.00	2.46	2.42	.04*

^*p<.05,

^**p<.01,

^***p<.001} one-tail

Table 5.

Mean Attitudes and Differences in Attitudes for the Family Formation Behaviors of Sons as Measured by Whether the Son Behavior Question Set Preceded or Followed the Daughter Behavior Question Set

Attitudes Towards Courtship	Asked About Son 1st	Asked About Son 2nd	Difference
Dating Young OK	4.41	3.64	0.77***
Sex Young OK	4.34	4.15	0.19
Cohabitation OK	3.12	2.80	0.32***
Timing of Marriage
Ideal Age at Marriage (Years)	26.85	26.15	0.70***
Work before Marriage Important	3.75	3.90	−0.15***
Marriage and Children
OK if No Marriage	2.88	2.74	0.14*
OK if No Children	2.75	2.62	0.13*

^*p <.05,

^**p< .01,

^***p <.001} two-tail

These questions were administered to each respondent twice, once pertaining to male children and once about female children. The respondents were randomly assigned to two groups – one group was asked about male children first and then about female children and the second group responded to questions about female children first. This random assignment experiment was designed to reveal the extent to which respondents altered their answers in the second series of questions depending on their answers to the first series of questions¹. Preliminary analyses of these data indicated that the measures were affected by the order in which they appeared in the protocol, that is, respondents tended to adjust the second set of responses based upon how they had answered the first set. Hence, a dichotomous measure was created to indicate which gender was asked about first – the daughter or son. The two sets of responses were analyzed as within family daughter/son comparisons as well as in pooled record formats where each respondent contributed a maximum of two records, one per gender of children.

Previous research demonstrates other factors such as age, education, marital status, parenthood and religion have strong effects on these attitudes toward courtship and family formation (Alwin, Cohen and Newcomb 1991; Cassidy and Warren 1996; Cunningham et al. 2005; Fan and Marini 2000; Gervai, Turner and Hinde 1995; Kiecolt and Acock 1988; Kinnaird and Gerrard 1986; Morgan and Waite 1987; Polit 1978; Roche 1986; Singh 1980; Stephan and Corder 1985; Thornton, Alwin and Camburn 1983; Trent and South 1992). With regard to sex and cohabitation, it has been well documented that women, less educated, older, and more religious persons in particular tend to have less permissive attitudes toward premarital sex and premarital cohabitation (Hendrick and Hendrick 1987; Hong 1983; Laner and Housker 1980; Maranell, Dodder and Mitchell 1970; Oliver and Hyde 1993; Sheeran et al. 1993; Sheeran et al. 1996; Singh 1980; Sprecher et al. 1988). Therefore our multivariate models also control for other aspects of respondents’ backgrounds. In the family formation domain we use dichotomous measures of the young adults’ own experiences with marriage (both current and previous), cohabitation (ever), parenthood (ever) and early sexual initiation (before age 18). For those who are parents our models sometimes include measures of whether or not the respondent has a daughter (1, 0), the age of the oldest child (in years), and the total number of children. In the socioeconomic domain we use educational attainment measured in years, the percent of time spent in full-time work during the last six months, and 1993 family income expressed in its natural log. Finally, in the religious domain we measure the respondent’s 1985 religious affiliation coded into a set of three dummy variables (Fundamentalist Protestant, Catholic, Jewish/Other/None), with Non-Fundamentalist Protestant serving as the reference group. We also measure the frequency of the respondent’s attendance at religious services with an ordinal variable coded 1 if the respondent never attends services, 2 if he/she attends less than once a month, 3 if the respondent attends once a month, 4 if he/she attends a few times a month, 5 if attendance occurs once a week, and 6 if the respondent attends several times a week or more. The univariate statistics for all these measures are presented in Table 2.

Results

Univariate Estimates of the Gender Double Standard

Our analysis begins with a simple comparison of attitudes toward sons’ and daughters’ courtship and family formation behavior. In order to make this comparison we computed a mean response for answers to each of the questions in the seven different subjects we investigate. We calculated the means separately for responses regarding sons and responses regarding daughters. Finally we compute the difference between the son and daughter means and use a simple difference of means test to determine the statistical significance of the gender differences in responses for each of the seven different subjects. Our sample includes both 31-year olds who are parents and 31-year olds who have never been parents, so we perform these calculations three times – once for the total sample, once for non-parents only, and once for parents only. Table 3 displays these results.

In the total sample (first column of Table 3) we find that respondents’ attitudes towards courtship behavior (dating, sex, cohabitation) and the timing of marriage (ideal age at marriage, work before marriage) differ significantly by the sex of the child. On average, respondents are more permissive toward a son dating (4.02), engaging in sex (4.24) and cohabiting (2.96) than they are toward a daughter participating in the same behaviors (3.43, 3.65, 2.71 respectively)². These differences are all statistically significant at a p-value of .001. On the other hand, the average ideal age at marriage for a son (26.49) is significantly higher than the average ideal age at marriage for a daughter (25.64) and the importance of work before marriage is also significantly higher for a son (3.82) than for a daughter (3.67). So parents and potential parents tolerate their sons engaging in courtship behavior earlier than their daughters, but prefer their daughters to marry earlier than their sons. Attitudes toward a son remaining single or childless are also somewhat more permissive than for a daughter, but only the difference in attitudes toward remaining single is statistically significant. This higher tolerance of sons remaining single appears consistent with the higher tolerance of sons marrying later.

In Column 2 of Table 3 we examine these same gender differences, but now only among those who have never been a parent. Surprisingly, the gender differences in attitudes about hypothetical sons and daughters among those who have never been parents are virtually the same as the gender differences in attitudes among those who are parents. These similarities are particularly strong in attitudes about courtship behaviors. Column 3 of Table 3 presents the gender differences in attitudes among only those who are parents, revealing that in the courtship domain the gender differences calculated in Column 2 are almost identical to those calculated in Column 3. These gender differences are also quite similar in the domain of the timing of marriage. Non-parents have a much larger gender difference in attitudes about their sons and daughters remaining single and parents have a much larger gender difference in attitudes about their sons and daughters remaining childless. So, although we predict a stronger gender difference in attitudes about the behavior of those who are socially close – one’s own children – we find within the domains of courtship and marriage timing virtually no differences between those who actually have children and those who have no children. In the domain of marriage and childbearing the contrast between parents and non-parents runs in opposite directions for specific measures. Note that among the parents their children are on average quite young, so that the courtship and family formation behaviors at the focus of these questions loom in the distant future. This may explain some of the similarities between parents and non-parents in this study.

Overall these findings are consistent with our expectations. The measures from this general population sample reveal substantial gender differences in parental attitudes toward their children’s dating, premarital sex, cohabitation, and marriage. The fact that these gender differences run in opposite directions for courtship and the timing of marriage is consistent with historical gender differences in the United States. The fact that attitudes toward children never producing their own offspring do not, overall, vary by gender of the child is consistent with our expectation that the gender double standard varies greatly across subject matter. Finally, the fact that the differences we observe for courtship and marriage timing are so strong is consistent with our focus on one’s own children – a group to whom the respondent has (or will have) extremely close social ties. This appears to be just as true for those answering about hypothetical future children as for those who already have children.

We also compared the size of the gender double standard across the topics represented by our measures. The metric of our measures varies greatly across topics, so we conducted this comparison by recalculating Table 3 for standard deviation scores, or z-scores. This comparison reveals the double standard in attitudes toward dating to be the largest, with a gender difference of more than a third of a standard deviation (.355). The double standard in early sex is .296 standard deviations, cohabitation is .221 standard deviations, age at marriage is .297, work before marriage is .265, not marrying is .058, and no children is .019. Similar to Table 3, results of this analysis also demonstrate much larger doubles standards for marriage timing and courtship than for not marrying or not having children. To estimate the statistical significance of these differences across domains, we then conducted difference-in-difference tests for each pair of domains. The results are presented in Table 4. Virtually every pair is significantly different, with two notable exceptions – the large gender double standard between attitudes about sons’ vs. daughters’ sex and dating is not significantly different and the very small (statistically insignificant) double standard between attitudes about sons’ and daughters’ not marrying and not having children is not statistically different. That is, within the domain of sex and dating and within the domain of not marrying or having children there is no significant difference in gender double standards. All the across domain comparisons reveal statistically significant differences.

Table 4.

Estimating Statistical Significance of Differences Across Family Formation Domains Using Standardized Difference-in-Difference Tests of Attitudes Toward Sons and Daughter Behaviors

	Courtship			Timing of Marriage		Marriage and Children
	Dating	Sex	Cohabit	Age at Mar	Work before Marriage	No Marriage	No Children
Courtship
Dating Young OK	---
Sex Young OK	0.10***	---
Cohabitation OK	0.04	.14***	---
Timing of Marriage
Ideal Age at Marriage (Years)	−0.58***	−0.68***	−0.53***	---
Work before Marriage Important	0.38***	0.48***	0.34***	−.20***	---
Marriage and Children
OK if No Marriage	−0.30***	−0.19***	−0.34***	−0.88***	−0.67***	---
OK if No Children	−0.27***	−0.18***	0.32***	−0.86***	0.66***	−0.02	---

^*p<.05,

^**p<.01,

^***p<.001} one-tail

Next we examine the effect of social desirability on responses. We accomplish this by focusing exclusively on answers about sons, but by comparing answers when respondents were asked about sons first to those when respondents were asked about sons second. The results in Table 5 reveal that when respondents are asked about the courtship behavior of sons after being asked about daughters (i.e., asked about sons second), they systematically adjust their responses about sons to be less permissive, making these responses more like their answers about daughters. For example, when asked about the timing of a son’s first dating, the mean among respondents who had not been asked about a daughter was 4.41 but the mean among those who had already been asked about a daughter was only 3.64 (first row of Table 5). On average respondents gave significantly less permissive answers about a daughter’s dating behavior (Table 3), and when asked about a daughter first they adjust their responses about a son to be more similar to their responses about a daughter. This difference is consistent with social desirability aimed at reflecting similar attitudes toward both sons and daughters, so that respondents give a different answer about their sons if they have already answered about a daughter than if they have not yet been asked about a daughter. The direction of this difference is the same for dating, premarital sex, and premarital cohabitation, but the differences were only statistically significant for dating and cohabitation.

Respondents who were asked about a son’s timing of marriage and a son not marrying or having children after being asked about a daughter also systematically adjusted their responses to be more similar to responses about daughters. All of these differences are statistically significant at a p-value of .05 or higher. Social desirability also affects responses about a son’s work before marriage, but here the effect is in the opposite direction. In this case, being asked about daughters first leads people to respond with answers about sons that are less similar to those about daughters. Although the direction of this effect was unexpected, it is consistent with concerns that social desirability may bias measures of the gender double standard. Overall these results indicate that the social desirability effect permeates a broader range of attitudes than the gender double standard estimated in Table 3.

Although one might expect symmetry in this social desirability, we actually find that respondents do not adjust their answers about their daughters to match answers about their sons in the same way that they adjusted their answers about their sons to match their answers about their daughters. Put another way, respondents’ restrictive attitudes toward daughters are not subject to the same level of social desirability effects as are their permissive attitudes toward their sons. These findings come from a comparison of those who were asked about daughters first to those who were asked about daughters second that is not displayed in the Tables. We find no statistically significant difference between the mean attitudes toward the courtship behavior or childbearing decisions of daughters (data not shown in tables). Difference in means for timing of marriage (−0.58 {ideal age} and 0.08 {importance of work}) and decisions to marry (0.21) were statistically significant (p-value of .05 or higher). Thus there is some evidence for social desirability effects among responses about daughters, but they are limited to the domain of marital behavior and do not include courtship or childbearing behavior. In the courtship domain respondents are more likely to hold fixed beliefs about what is considered appropriate behavior for daughters and tend to be more flexible about their attitudes toward sons. We find a similar result for attitudes toward remaining childless. Below we examine multivariate models to adjust our estimates of the gender double standard for this social desirability effect.

Multivariate Models of Parents’ Attitudes toward Their Children’s Behavior

We now turn to estimates from multivariate models that control for the social desirability effect while estimating effects on parental attitudes about children’s behavior, including the influence of the gender of the child. These data have been transformed so that each respondent contributes two observations to the analyses; one record speaks to attitudes about sons while the other addresses attitudes about daughters. We analyze multiple records per respondent, and records from the same respondent are likely to be similar, so we estimate these multivariate models using multilevel estimation techniques (two reports clustered within a single individual)³. The two-level models allow us to make appropriate adjustments for the error variance correlation between the two reports from the same respondent (Raudenbush and Bryk 2002). As discussed above, we have good reason to expect that a number of other personal characteristics and previous experiences influence individuals’ attitudes about each of these courtship and family formation behaviors. Therefore in these multivariate models we also add measures of these characteristics and experiences as controls. Our results are displayed in Table 6.

Table 6.

Two-Level Models Predicting Respondent Attitudes Toward Children’s Behaviors Using Pooled Daughter and Son Measures (t-ratios in parentheses)

	Courtship			Timing of Marriage		Marriage and Children
	Dating Young OK	Sex Young OK	Cohabit OK	Ideal Age (Years)	Work Before Important	OK if No Marriage	OK if No Children
Child’s Gender
Son	.58*** (13.29)	.58*** (12.85)	.25*** (10.09)	.89*** (15.28)	.16*** (9.28)	.07** (2.59)	.03 (1.27)
CONTROLS
Social Desirability
If Daughter Asked 1st	−.32*** (3.18)	−.04 (.34)	−.13* (2.09)	−.53*** (3.20)	.10** (2.85)	.04 (.57)	−.04 (.58)
Respondent Gender
If Respondent Male	.46*** (4.17)	.36** (2.80)	−.05 (.74)	−1.30*** (7.00)	−.17*** (4.17)	.18** (2.42)	.06 (.84)
Respondent Family Formation Experience
Currently Married	−.16 (1.02)	−.62*** (3.52)	−.31*** (3.19)	−1.15*** (4.49)	−.03 (.56)	−.16 (1.53)	−.24** (2.38)
Previously Married	.12 (.60)	.14 (.59)	−.03 (.20)	−.49 (1.44)	−.09 (1.25)	.03 (.20)	−.15 (1.16)
Ever Cohabited	.21* (1.94)	.64*** (5.12)	.69*** (9.97)	.55** (3.03)	.01 (.38)	.02 (.31)	.17** (2.39)
Ever a Parent	−.33** (2.55)	−.21 (1.41)	−.20** (2.46)	−.71*** (3.32)	−.01 (.31)	−.19* (2.17)	−.39*** (4.62)
Had Sex by Age 18	.10 (.88)	.43*** (3.31)	.05 (.74)	.29 (1.56)	.03 (.83)	.04 (.52)	.04 (.58)
Respondent Education, Work, Income
Education (years)	.10*** (3.63)	−.01 (.33)	.002 (.10)	.22*** (4.79)	−.01 (1.28)	−.01 (.33)	.01 (.36)
Work (% last 6 months full-time)	−.002 (1.48)	.002 (.97)	.002 (1.52)	.002 (.55)	.000 (.08)	.002* (2.18)	.003** (2.99)
Income (log 1993 family income)	.04 (.88)	.09* (1.90)	.04 (1.35)	.02 (.33)	.01 (.44)	.01 (.42)	−.01 (.52)
Respondent Religion
Fundamental Protestant	−.31 (1.63)	−.35 (1.63)	−.21* (1.74)	−.61* (1.93)	.04 (.67)	−.14 (1.09)	−.17 (1.42)
Catholic	−.29** (2.37)	−.14 (.95)	.01 (.17)	.20 (.97)	.10* (2.21)	−.13 (1.53)	−.13 (1.61)
Jewish, Other, None	−.01 (.05)	..06 (.27)	−.06 (.55)	.73** (2.33)	.04 (.54)	−.14 (1.17)	−.17 (1.42)
Religious Service Attendance	−.04 (.90)	−.27*** (6.11)	−.17*** (6.98)	−.04 (1.56)	−.02 (1.52)	−.05* (1.86)	−.03 (1.12)
N	1658	1654	1657	1652	1661	1659	1658

^*p<.05,

^**p<.01,

^***p<.001} one-tail

Each column in Table 6 displays the results for a model of each specific attitude toward courtship, marriage timing, and remaining unmarried and childless. Coefficients in row 1 of Table 6 reflect fully controlled child gender effects on respondent attitudes toward courtship and family formation behaviors. These gender effects remain consistent and statistically significant even after the estimates control for social desirability as well as other known predictors of these attitudes⁴. Respondents continue to be significantly more receptive to earlier ages at dating and sexual initiation for sons, and more accepting of sons’ cohabitation, older ages at marriage, working before marriage, and remaining single. Differences in responses concerning the acceptability of not having children remain non-significant. All of these estimates are completely consistent with the bivariate results presented in Table 3⁵. All in all, these findings provide strong evidence of a substantial gender double standard in parenting attitudes toward children’s courtship and marriage behavior, although the extent of this double standard varies across topics.

Table 6 controls reveal important insights as well. Note that this formulation of the model focuses on the distribution of parents’ answers about their children, not the gender difference in those answers. As discussed in the Table 5 social desirability analyses, when questions referring to daughter dating or cohabitation behaviors are administered first, responses are significantly less accepting than if the questions are asked after the respondent has answered about sons. Social desirability also has significant effects on attitudes about ideal age at marriage and working full-time before marriage. One of the most interesting aspects to note is that, in this formulation of the model, the order in which it is asked does little to influence what parents say about children’s age at sexual initiation. Considering the strong positive effects of “if son record” on this attitude, the willingness to display candor provides some evidence for the argument that an open sexual double standard exists in the domain of sexual behavior. At least in this formulation of the model, which focuses on parents attitudes not the gender difference in those attitudes, parental attitudes toward sex are not affected greatly by social desirability.

Analyses displayed in Table 6 verify the mostly significant effects of parental gender on endorsements of child courtship and family formation behaviors. Men report acceptance for younger ages at dating as well as younger ages at sexual initiation, and are less bothered than women if their offspring do not marry. Men also, on average, report a lower ideal age for marriage and are less likely to feel strongly about their children working full-time before marriage.

A respondent’s own experiences with marriage and family formation also affect attitudes towards children’s courtship and marriage behaviors. Currently married people or people who have already had children are less accepting of younger ages at dating, younger ages at initiation of sexual activity, and cohabitation without marriage. As one might expect, respondents who have cohabited without marriage are significantly more accepting of those same behaviors for their children. Anyone with current marital or parental experience endorses a younger ideal age at marriage, while those who have cohabited express an older ideal age. Currently married respondents or those who have had children have significantly less tolerance for their children not having their own offspring. Being a parent has significant negative effects on acceptance for children not marrying. People who have had cohabitation experience indicate that they are more accepting of their children choosing not to become parents. Finally, those who had sex by age 18 are significantly more accepting of their children engaging in premarital sex.

Education, work, and income play roles in shaping some of the attitudes within the courtship, marriage, and childbearing domains. Those with more years of formal education endorse younger ages at dating and are proponents of older ages at marriage. These findings are both statistically significant. Higher percentages of time spent in full-time work activities are associated with slightly accepting attitudes toward remaining single and remaining childless. Those with higher income also show somewhat more accepting attitudes toward premarital sex.

Respondents’ religious affiliations also have a significant effect on their attitudes toward their children’s courtship and family formation behaviors. Fundamentalist Protestants are somewhat more restrictive in their views concerning age at dating and sexual initiation than non-Fundamentalist Protestants, and significantly less accepting of cohabitation than non-Fundamentalist Protestants. Fundamentalists also prefer younger ideal ages at marriage than non-Fundamentalists. On average Catholics are significantly less accepting of early dating and place a higher importance on working before marriage than non-Fundamentalist Protestants. Those classified in the Jewish/Other/None category endorse significantly older ages at marriage than non-Fundamentalist Protestants.

Religiosity, as measured by attendance at religious services, has particularly strong effects on attitudes about children’s sexual and cohabiting behavior. Respondents with a high level of religious service attendance are significantly less likely to approve of either early sexual behavior or premarital cohabitation. More religious respondents are also somewhat less likely to be tolerant of their child deciding not to marry. Religiosity did not have a significant effect on the other courtship and family formation attitudes.

We also estimated these same models among the more restricted sub-sample of only those respondents who actually had children of their own. This restriction allowed us to investigate the influence of characteristics of the children on these same parenting attitudes. The results of this investigation are presented in Table 7. Just as in the bivariate results presented in Table 3, the effects of the child’s gender are virtually the same in the parent-only sub-sample (Table 7) as they were in the total sample (Table 6). Important differences limit any comparison of results in Tables 6 and Tables 7 – including both the smaller sample size and added explanatory factors in the models displayed in Table 7 – nonetheless, many of the other significant effects displayed in Table 6 are also replicated in Table 7. The biggest differences involve the addition of measures of childbearing experiences among these parents. Parents with daughters report significantly less tolerance of early dating than parents with no daughters (row 7, column 1, Table 7). Parents with older children reported lower ideal ages for marriage than parents with younger children (row 8 of Table 7). All the parents in this study were age 31 at the time of the interview, so those with older children themselves began family formation at younger ages than the other parents, which may explain this result. Parents with older children also reported significantly higher importance of working before marriage than parents with younger children. Parents with many children reported lower ideal ages at marriage and less importance of work before marriage. Clearly characteristics of these 31-year old parents’ actual children do shape their responses to some of these questions about parenting attitudes.

Table 7.

Two-Level Models Predicting Respondent Attitudes Toward Children’s Behaviors Using Pooled Daughter and Son Measures (t-ratios in parentheses) – Parents only

	Courtship			Timing of Marriage		Marriage and Children
	Dating Young OK	Sex Young OK	Cohabit OK	Ideal Age (Years)	Work Before Important	OK if No Marriage	OK if No Children
Child’s Gender
If Son Record	.59*** (10.15)	.60*** (10.19)	.29*** (8.99)	.75*** (11.35)	.16*** (6.89)	.01 (.18)	.04* (1.67)
CONTROLS
Social Desirability
If Daughter Asked 1st	−.19 (1.52)	.18 (1.25)	−.03 (.34)	−.71*** (3.54)	.11** (2.51)	.03 (.33)	−.08 (.95)
Respondent Gender
If Respondent Male	.33* (2.20)	.36* (1.99)	−.11 (1.01)	−1.51*** (6.13)	−.15** (2.86)	.33*** (3.09)	.13 (1.22)
Respondent Family Formation Experience
Currently Married	.00 (0.00)	−.67* (1.65)	−.23 (.99)	.17 (.30)	.02 (.21)	.10 (.42)	.00 (.02)
Previously Married	.01 (.01)	−.25 (.57)	−.13 (.49)	.38 (.62)	−.03 (.22)	.30 (1.15)	−.09 (.36)
Ever Cohabited	.21 (1.54)	.50** (2.98)	.66*** (6.77)	.54** (2.37)	−.03 (.62)	−.02 (.20)	.19* (1.98)
Has Daughters	−.37** (2.76)	−.11 (.68)	.01 (.09)	.33 (1.47)	.00 (.08)	.01 (.10)	.03 (.32)
Age of Oldest Child	−.01 (.34)	.02 (.81)	.01 (.77)	−.08* (2.17)	.02** (2.37)	.02 (1.11)	.01 (.79)
# of Children Ever Born Or Adopted	−.02 (.28)	−.02 (.15)	−.07 (1.20)	−.27* (1.84)	−.09** (2.93)	.05 (.82)	.02 (.40)
Had Sex by Age 18	.12 (.83)	.52** (2.97)	.14 (1.38)	.25 (1.05)	.01 (.17)	.06 (.54)	.09 (.91)
Respondent Education, Work, Income
Education (years)	.03 (.87)	−.08* (1.86)	−.04 (1.57)	.08 (1.26)	−.02 (1.22)	−.01 (.39)	.00 (.17)
Work (% last 6 months full-time)	−.001 (.56)	.003 (1.23)	.002 (1.52)	.01 (1.53)	.00 (.96)	.002* (1.70)	.003** (2.33)
Income (log 1993 family income)	.09* (1.75)	.10 (1.60)	.05 (1.36)	−.10 (1.10)	.02 (1.24)	.03 (.88)	.00 (.07)
Respondent Religion
Fundamental Protestant	−.20 (.91)	−.28 (1.06)	−.23 (1.51)	−.57 (1.60)	.06 (.75)	−.30* (1.97)	−.21 (1.34)
Catholic	−.26* (1.70)	.02 (.10)	.05 (.48)	.48* (1.91)	.13** (2.40)	−.09 (.86)	−.09 (.85)
Jewish, Other, None	−.09 (.38)	.09 (.31)	−.09 (.53)	.56 (1.37)	−.04 (.44)	−.30* (1.69)	−.21 (1.19)
Religious Service Attendance	.01 (.26)	−.22*** (3.76)	−.14*** (4.07)	−.08 (1.05)	−.02 (1.05)	−.06* (1.84)	−.02 (.68)
N	993	989	993	990	994	994	993

^*p<.05,

^**p<.01,

^***p<.001 } one-tail

Conclusion

The evidence we present here is consistent with the conclusion that there is a substantial gender double standard in attitudes about children’s courtship and family formation in the adult population. By focusing on attitudes toward courtship and family formation behavior among those with extremely close social ties – one’s own children – our measures reveal a considerable gender double standard. Our experimental design reveals important consequences of social desirability for measurement of the gender double standard, but in this substantive domain, in reference to one’s own children, the double standard is strong enough to be observed even without controlling for social desirability effects. Controlling for social desirability, the gender double standard is large and robust even in multivariate models controlling for many other known determinants of these attitudes. We conclude this gender double standard in parenting values is substantial and deserves careful research attention.

Research on the gender double standard will confront important obstacles. Our investigation demonstrates that both the magnitude and the direction of these double standards vary by subject matter. Even though we focus on subjects of courtship and family formation, the direction of this double standard varies across specific domains of courtship and family formation, with people preferring later courtship but earlier marriage for their daughters. Given these differences within the courtship and family formation domain investigators will need to be vigilant for substantial variation in the magnitude and direction of the gender double standard across other domains. Likewise, social desirability also appears to affect measurement of this double standard, so that failure to mitigate the social desirability effect may produce under-estimates of the true gender double standard. As a result any research on gender double standards will need to give careful attention to the possible effects of social desirability.

These double standards in parents’ values and preferences for their sons and daughters may hold one key to understanding persistent gender differences in courtship and family behavior. For example, here we document a tendency to hold significantly more permissive attitudes toward a future son’s courtship experiences than a future daughter’s courtship experiences. We find this substantial gender double standard in the 1990s, in spite of three previous decades of social change toward more permissive attitudes about sex and more egalitarian attitudes toward men’s and women’s roles (Thornton and Young-DeMarco 2001; Veroff et al. 1981). It appears that these gender double standards in parents’ preferences for their children’s behavior have not been eradicated by these other changes toward more permissive and egalitarian attitudes. Waite and Goldscheider (1991) argue that marriage and childbearing in the United States will either continue to decline or be fueled by a second gender role revolution – continued social change toward egalitarian gender role attitudes that include more widespread endorsement of men’s involvement in family life. Persistent gender double standards about sex and courtship are not the same as double standards about men’s roles in families. However, the persistence of parenting gender double standards likely contributes to the persistence of separate and distinct gender roles. As long as substantial double standards remain a part of parental socialization of children, completely egalitarian gender role attitudes seem unlikely, and the gender gaps in family behaviors are likely to remain.

Further research on gender double standards may hold one of the keys to understanding the origins of gender differences in other domains of social behavior as well. For example, here we document parents’ tendencies to prefer older ages of marriage for their sons than for their daughters. These double standards parallel some of the observed behavioral differences in the general population – women enter marital relationships earlier than men (Casper and Bianchi 2002; Thornton et al. 2007; Waite and Goldscheider 1991). Additional research will be needed to illustrate the point empirically, but it seems quite likely that these gender double standards in parenting values shape, at least in part, the important gender differences in family formation behavior. If true in this domain, then parents’ gender double standards in other domains of social life may also contribute to gender differences in behavior. Given the strong double standards documented in the parenting attitudes examined here, the intergenerational origins of the gender double standard may prove particularly fruitful. This line of inquiry is a high priority to investigate the factors shaping parenting double standards that may ultimately produce substantial gender differences in social behaviors.