Family Formation and Men's and Women's Attainment of Workplace Authority

An equal gender distribution in authority in the workplace is important not only because authority positions reward the individual with prestige, power, autonomy and status, but because male dominance in positions of authority may also be a significant cause of gender inequality in other dimensions of the labor market (Wright, Baxter and Birkelund 1995). Studies indicate that the more male-dominated the managerial staff, the larger the gender wage gap among subordinates (Cohen and Huffman 2007; Hultin and Szulkin 2003) and the lower the chances of promotion for subordinate women (Cohen, Broschak and Haveman 1998). The gender gap in authority also accounts for part of the gender wage gap (England et al. 1994; Halaby and Weaklim 1993; McGuire and Reskin 1993). In other words, more women in positions of authority could contribute to gender equality more generally. Another argument for increased gender equality, although less often mentioned, is that if women – for different reasons – choose not to strive for power and authority, or if they are systematically debarred from these positions, society as a whole suffers; it misses the unexploited talent that these women possess.

Studies on the gender gap in authority have exclusively been based on non-repeated cross-sectional (and predominantly U.S.) data. Hence our knowledge about this phenomenon is limited. Although studies have uniformly shown that women have less authority than men do, we do not know much about whether the size of the gap has changed over time or about the mechanisms governing the process by which individuals move into and out of authority positions.

In this article we use cross-sectional and panel data from the Swedish Level of Living Survey collected in 1968, 1974, 1981, 1991 and 2000, to contribute to research aiming at describing and explaining the gender gap in authority. The basic research questions we wish to answer are, first, how has the gross gender gap in authority changed over time during the past decades and, second, to what extent can the gap be attributed to gender-specific consequences of family formation (i.e., marriage, cohabitation and parenthood)? It is a well-known fact that women in all societies take a larger responsibility for children and household chores. Relatively little is still known, however, about the impact of these responsibilities on women's chances of gaining authority in the labor market. The results of the few studies that have been conducted are inconclusive and methodologically problematic. As pointed out in a review by Smith (2002), additional research in this area is greatly needed to specify the relationship between these responsibilities and the attainment of workplace authority. The panel structure of the data enables us to take a step toward a causal understanding of the relationship between individuals' family situations and their career outcomes. Building on insights from family sociology, sociology of work and labor economics, our study is the first to account for selection effects due to unobserved time-invariant heterogeneity. Thus, our results give an indication of whether, and to what extent, previous results, based on cross-sectional data, are biased because of non-random selection into family formation.

Mechanisms Behind the Gender Gap in Workplace Authority

Researchers have defined and operationalized the concept of workplace authority in numerous ways, for example as the right to hire and fire, set the rate of pay, supervise the work of subordinates, participate directly in policy decisions in the workplace, and the possession of a formal position of authority (like an executive title) (McGuire and Reskin 1993; Wolf and Fligstein 1979; Wright et al. 1995).¹ A common characteristic for these dimensions, however, seems to involve “legitimated control over the work process of others.”(Wolf and Fligstein 1979:236) Regardless of definition, previous studies have uniformly shown that women are less likely to exert workplace authority than are men (Smith 2002). Moreover, when women are in positions of authority, they are located at lower levels of management. Thus they exert less authority than do male managers (Hopcroft 1996; Jacobs 1992; Jaffee 1989). In accordance with most previous research we study supervisory authority (i.e., whether the respondent has subordinates) in the empirical part of this article.²

Family Formation

A large amount of evidence indicates that there are gender-specific ways of reconciling family and career-related demands (e.g., Blossfeld and Drobnic 2001; Hakim 2000). One consequence of the gender division of labor is that women (mothers) commonly have less work experience than do men (Edin and Richardson 2002; Hultin 1998; Mueller, Kuruvilla and Iverson 1994). Because the prospects of reaching positions of power, influence and authority are conditioned by the individual's human capital, such differences usually explain some of the gender gap in authority. Studies, however, consistently reveal a remaining and pronounced authority gap between women and men even after controlling for various indicators of human capital, across countries (Baxter and Wright 2000 [Australia, Sweden and the United States]; Huffman and Cohen 2004 [USA]; Hultin 1998 [Sweden]; Jaffee 1989 [USA]; Kraus and Yonay 2000 [Israel]; Mitra 2003 [USA]; Wright et al. 1995 [seven Western countries]).³

Over and above human capital differences, family responsibilities might affect women's authority attainment in other ways as well. In this context, status characteristics involving gender and parental status may be influential. A status characteristic is activated, or becomes salient, when it differentiates actors or is assumed to be relevant in a particular context. Individuals then use the status characteristic to assign expectations – to behavior, competence, commitment, etc. – that are in accordance with their status beliefs (Wagner and Berger 2002). Status characteristics appear to be rather blatant in the evaluation of mothers' and fathers' parenting skills and job-related competence. Parental status has been shown to have a polarizing effect on judgments of individual men and women in evaluations of job competence, and decisions on employment and promotion. Several U.S.-based experimental studies have shown that being a parent activates relatively harsh job-related standards for a woman, but relatively lenient standards for a man, indicating the presence of a status-based discrimination mechanism, to the disadvantage of mothers and the advantage of fathers (Cuddy, Fiske and Glick 2004; Correll, Benard and In Paik 2007; Fuegen et al. 2004).

Whereas discrimination in status-based theories derives from a cognitive bias operating over and above the actual productivity of individuals, discrimination in theories of statistical discrimination derives from informational bias; employers, lacking information on individual productivity, are thought to base their decisions on information of average differences in the productivity of groups (cf., Correll and Benard 2006). If employers assume or know that women, on average, are more constrained by family responsibilities than are men, and/or leave the labor market for longer periods, they may statistically discriminate using this information in their employment and promotion decisions. This might specifically be the case in societies such as Sweden, characterized by family-friendly policies (i.e., long parental leave periods and/or institutionalized options for parents to work parttime) which, due to the traditional gender division of labor, are in practice mainly directed at mothers (cf., Mandel and Semyonov 2005). High turnover rates in positions of authority can also be assumed to be especially costly for an employer, and to avoid these costs, they may avoid hiring married and cohabiting women and/or women “at risk” of having children, or mothers of small children, for these positions (cf., Bielby and Baron 1986).

As dual-earner couples have become increasingly common, the traditional male breadwinner model has lost significance: men and women increasingly share their unpaid household work and some men reduce their time for market work following childbirth (cf., Bygren, Gähler and Nermo 2004). Nevertheless, the male “good-provider model” prevails in the sense that men most commonly continue working fulltime or even increase their market work hours when they become fathers, whereas women often decrease their labor supply when they become mothers (Kaufman and Uhlenberg 2000 [USA]; Kennerberg 2007 [Sweden]). While Sweden is often acclaimed for its gender equality and its gender egalitarian attitudes (e.g., Inglehart 2003; Svallfors 2006), there is ample evidence that Swedish women and men still act in accordance with traditional norms after union formation and childbirth. Mothers use a large majority of parental leave days. After parental leave mothers dedicate more time to childcare (Statistics Sweden 2003) and often work reduced hours for long periods, whereas fathers usually continue working fulltime (Kennerberg 2007; Sundström 1997). The gender difference in time devoted to household tasks seems to be triggered by marriage/cohabitation and childbearing: It is relatively small for single persons, but grows substantially larger among childless couples and further increases when couples have children (Statistics Sweden 2003). Not only does parenthood change behaviors; parenthood appears to make Swedish men and women more traditional also with regard to their attitudes to the gender division of labor in the family (Bernhardt 2005), and entry into motherhood has been found to have a causal effect lowering Swedish women's work commitment (Evertsson and Breen 2008). Further, qualitative studies indicate that family responsibilities are regarded as intrinsic to motherhood and defined as problematic to reconcile with job demands in Swedish workplaces. For fathers, care responsibilities are most often understood as optional, and from the perspective of employers and co-workers, fatherhood is believed to make men more mature and even better as managers (Bekkengen 2002; Kugelberg 2006).

The scant available empirical evidence also suggests that parenthood is negative for women's chances to reach positions of authority, and positive for men's chances, but an absence of association between parenthood and authority is actually the most common finding. A weakness with studies on the attainment of workplace authority, however, is that they exclusively utilize non-repeated cross-sectional data, so the potential problem of selection bias in the estimated effects of parenthood and/or civil status cannot be dealt with in a serious way. Bearing this in mind, Hultin (1998) finds that Swedish women with children have lower chances of exerting workplace authority compared to childless women, but no such child penalty is found for men. Rosenfeld, van Buren and Kalleberg (1998) (using a nine-country dataset) and Wolf and Fliegstein (1979) (using two U.S. samples) report equal results for women, but in their studies men with children benefit instead. In contrast to these results, more recent U.S. studies find no effects of children on authority for either men or women (see Hopcroft 1996; Huffman and Cohen 2004; Mitra 2003). Comparing the authority outcomes of married/cohabiting working men and women to those of single working men and women, the results are also mixed. Hultin (1998) finds that Swedish men who are married or cohabiting experience an authority premium, but that civil status is unrelated to workplace authority among Swedish women. Rosenfeld et al. (1998) find a marriage/cohabitation authority premium, but that it is given to both men and women. By contrast, Huffman and Cohen (2004) and Mitra (2003) (both using U.S. data) find that civil status is unrelated to authority for men and women alike.⁴

On the basis of this review, our empirical expectation is that cohabitation, marriage and parenthood have divergent effects for Swedish men and women. Because cohabitation, marriage, and most of all parenthood, go hand in hand with gender-specific changes in labor market-related inputs, household and family responsibilities, as well as employer and co-worker expectations around such changes, we expect that cohabitation, marriage and parenthood should increase men's chances, but decrease women's chances of attaining workplace authority. Concerning selection biases in cross-sectional estimates of the effects of these family related-factors, our guess is that they play a role. Unobserved factors are likely to influence mating, fertility and labor market outcomes, but we regard the directions of biases as an empirical question.

Context: Sweden Over Time

As noted, our study provides new evidence on these matters in the context of Sweden, which consistently ranks at the top in international comparisons of the degree of gender equality (e.g., United Nations Development Programme 2005). Swedish women have a higher labor market participation rate than women in most other countries (Jaumotte 2003), and the wage gap between working men and women is low in an international perspective (Blau and Kahn 2003). Furthermore, Swedish women are not economically dependent on their husbands to the same extent as are their counterparts in other countries (Evertsson and Nermo 2004; Sørensen 1994), and Swedish couples have a relatively egalitarian division of household work (Fuwa and Cohen 2007).

Yet, there remain substantial inequalities between Swedish men and women, and one of the more pronounced is the gender gap in workplace authority. Few women are found at the top of business firms, and female employees less often have managerial positions and subordinates (Hultin 1998; Mueller et al. 1994). In fact, the gap seems to be relatively large in Sweden compared to that in many other developed countries (Baxter and Wright 2000; Mandel and Semyonov 2006; Rosenfeld et al. 1998; Wright et al. 1995; Yaish and Stier 2009), even though it has been an explicit policy goal that more women should fill managerial positions (e.g., Ministry of Industry, Employment and Communications 2003).

The little we know about the development of the gender gap in authority over time is that women have increased their share of authority positions, but that the gap remains large. For Sweden, Meyersson Milgrom and Petersen (2006) show that the gender gap in “rank” (i.e., a job's level of difficulty and responsibility among other things) narrowed during the period 1970-1990. Studies using U.S. data indicate that more women in past decades have become managers (Reskin and Padavic 1994; Stainback and Tomaskovic-Devey 2009). While progress seems to have stalled in the 1990s (Cohen, Huffman and Knauer 2009), this increase was associated with women possessing real supervisory authority rather than just being assigned managerial titles (Jacobs 1992).

Since the 1970s, gender inequality has decreased in a number of domains: education, employment rates, work hours, wages, economic dependency within couples, household work hours and the use of parental leave (Bygren et al. 2004). Sweden has implemented a number of family-related policies aimed at replacing the male-breadwinner model with that of a dual-earner model. Important elements driving this transformation were the introduction of individual taxation of married couples in 1971, the introduction of parental-leave benefits in 1974, and the expansion of subsidized public childcare during the 1970s and 1980s (e.g., Lewis and Åström 1992). Although major inequalities remain, Swedish women of today should meet the labor market with more assets and possibilities relative to men than did women just a couple of decades ago.

As noted, though, the gender gap in workplace authority and access to managerial positions is actually relatively large in Sweden, and larger than in the United States (Mandel and Semyonov 2006; Rosenfeld et al. 1998; Wright et al. 1995). It has been hypothesized that the relatively large Swedish gender gap in workplace authority is partly caused by the relatively few managerial positions compared to the American labor market (Wright et al. 1995). A contributing factor might also be gender segregation by occupation (cf., Huffman and Cohen 2004), which is relatively high in Sweden, at least compared to the United States (Nermo 2000). The extensive parental leave program in Sweden has also been pointed out as a potential cause (see, e.g., Hakim 2000; Mandel and Semyonov 2006; Rosenfeld et al. 1998; Wright et al. 1995). In practice, a large share of Swedish mothers have a relatively weak connection to the labor market in the period following childbirth, due to long periods of absence and/or part-time work. Whereas most American mothers have returned to work within six months after the birth of a first child, a majority of Swedish first-time mothers have still not returned to work after two years (Aisenbrey, Evertsson and Grunow 2009). As employers normally prefer to promote (at least) full-time employees for authority positions, most Swedish women with small children may effectively be sorted out of the pool of employees considered suitable for these kinds of positions. Absence and part-time work among mothers may also give rise to statistical discrimination of female employees during fertile ages (Mandel and Semyonov 2006). Further, due to extensions of the parental leave program, the length of parental leave periods for women has increased during recent decades.⁵

It is important to bear in mind that a one-country study such as this one cannot speak to or resolve the role of the dual-earner model for the gender gap in workplace authority. Although our data cover a period before and after the implementation of these policies, we cannot analytically separate the effects of different dual-earner policies (for example, the public provision of childcare may promote gender equality, whereas the right to take long parental leaves may not), and we cannot distinguish between the effects of these policies from those of other causes that vary over time and are unobserved in our data (e.g., changes in norms, politically driven changes such as the use of gender quotas in the public sector).

Contribution of the Present Study

One aim of this study is to fill a void in the literature by supplying data on the historical development of workplace authority. A more serious void in this line of research, however, is the fact that longitudinal panel data have not been used previously. As a consequence, the issue of unobserved selection effects has routinely been ignored. Selection issues have been extensively dealt with in studies of the male marriage wage premium and the motherhood penalty on wages, but in studies of workplace authority, none has hitherto used panel designs. All of what we know about the attainment of authority is based on data from cross-sections, which seems far from ideal if we wish to approach a causal understanding of the authority attainment process. A well-known weakness of cross-sectional data in this context is that the effect of unobserved individual characteristics, which are correlated with observed independent variables, may give rise to omitted variable bias in the estimated coefficients (e.g., Halaby 2004). For example, a motherhood penalty on authority in a cross-section may have arisen because employers have used traits unobserved in the data to sort women into positions of authority. If these traits are correlated with the probability of being a mother, omitted variable bias in the estimated effect of motherhood would arise.

In the present study, we have access to an individual-level panel dataset covering a period of more than 30 years. Because many individuals in this dataset have been observed at more than one point in time, we are able to account in our estimations for time-invariant characteristics that sort individuals into authority positions. Our basic analytical strategy is to estimate fixed-effects regression models accounting for individual observed and unobserved heterogeneity and compare these to standard regression models accounting for observed heterogeneity but ignoring individual unobserved heterogeneity. In this way we are able to give an indication of whether, and if so to what extent, previous results, based on cross-sectional data, are biased because of non-random selection into family formation, and approach a causal understanding of the relationship between individuals' family situations and their career outcomes.

Data and Variables

For the analyses, we use cross-sectional and panel data from the Swedish Level of Living Survey collected in 1968, 1974, 1981, 1991 and 2000.⁶ The respondents in each cross-section constitute a .1 percent random sample of the Swedish adult population ages 18-75. The unique strength of these data is that they include detailed and longitudinally identical measures of labor market and family-related variables.

To enable our empirical analysis, we excluded respondent-years when a respondent did not work at least one hour a week.⁷ Additionally, we excluded a small number of observations with internal non-responses on relevant questions, and respondents past the age of 65. After these selections, the samples consisted of 2,816 respondents in 1968, 3,104 respondents in 1974, 3,328 respondents in 1981, 3,333 respondents in 1991, and 2,998 respondents in 2000 (which equals 15,579 observations in total). Due to the panel structure, the data contain information from 7,026 unique respondents, of which 63 percent participated in more than one cross-section of the survey. Descriptive statistics are presented in Table 1.

In each survey, all employed respondents were asked identical questions about whether they currently had any supervisory tasks. Respondents with supervisory tasks were also asked how many subordinates they had. We conducted a number of alternative analyses, using various cutting points and categorizations of the authority measure. Independent of measure and link function used (logits, tobit, OLS), the results from these analyses were very similar.⁸ For reasons of simplicity we therefore decided to use a dichotomized measure of workplace authority (i.e., whether or not the respondent has any subordinates). For the independent variable woman, we assigned the value 0 for males and 1 for females. We constructed three dummy variables indicating various dimensions of the respondent's children. The first indicates the presence of one or more preschool-aged children (0-6 years) in the household. The second indicates the presence of one or more children ages 7-20 in the household (7-18 years in the 1991 and 2000 surveys). The third indicates whether the respondent has grown up children (20+) who live in or outside the respondent's household (18+ years in the 1991 and 2000 surveys). We used a dichotomous measure of marital status with a value of 1 if respondent is married or cohabiting, and a value of 0 if not. We measured education in years of formal education: “For how many years have you been in full-time school and vocational education (from first grade and onwards)?” To study the impact of labor market conditions, we used four different indicators. We measured work experience as the respondent's total number of years in employment: “How many years altogether have you spent in gainful employment?” We measured seniority as the number of years since the respondent was first employed by the current employer. It should be kept in mind that this variable most probably is an overestimation of actual seniority, and more so for women. The measure of hours of paid work per week refers to the question “How many hours per week do you regularly work?” Finally, we used a variable of sector of employment, private, with a value of 1 if the individual is employed by a privately owned firm and 0 otherwise.

Findings

In Figure 1, we display the odds of a woman to be in a position with authority, relative to the share of women working, and the probability that an employee with authority is a woman, for the different years. A number of patterns are noteworthy. First, consistent with all other studies in this area of research, we find that women less often than men execute supervisory tasks. Second, during the period, the odds of women exerting authority increased substantially between 1968 and 1981 and, to a smaller extent, between 1981 and 2000. Hence, during the 1970s, there was a clear change in the direction toward more gender equality in authority outcomes, but this development leveled out at the beginning of the 1980s. The odds hold constant the share of women working. However, one might also ask what the probability is that a position of authority is held by a woman in a given year. We can see that this probability increased from 24 percent in 1968 to 39 percent in 2000. This is a combined effect of two changes that mainly took place in the 1970s: that of women increasing their relative share of the labor force and that of a gender equalization of chances (odds) of getting into authority positions.

In Figure 2, we display the proportions having at least one subordinate for different age groups, separately for men and women, pooled over survey years. Men and women start out at the same low level in their 20s, but men have a much steeper “career curve,” peaking in their late 40s, declining thereafter. Women's chances of having authority increase until they reach their 30s, but thereafter remain at a more or less flat level, significantly below that of men. Hence, men's and women's career development does not differ a great deal until they reach their 30s, when their paths diverge. That is, if there is a glass ceiling to women's careers, they hit it in their 30s. This point in life is also the age at which many Swedish men and women are in the stage of forming families (Statistics Sweden 2008). The question that remains to be answered is whether these gender-specific career profiles are associated with family-related events.

Panel Analysis

In effect, we thereby capture all observed and unobserved individual characteristics that do not change over the observation period, as well as all observed and unobserved gender-specific factors that for a particular survey year affect the probability of being transferred into or out of a position of authority (because the models are estimated by sex). Our point in estimating this model is to show what happens to the dependent variable when individuals experience a family-related change, for men and women separately. Keep in mind that the fixed-effects model uses only intra-individual variation to estimate effects, which means that these are only representative of the subselection(s) of individuals exhibiting such variation. Out of the sample of 7,026 individuals, 905 men and 599 women switch between having and not having authority during the period. A minority of the sample thus contributes to the estimation of these effects.

For all analyses, we introduced the family-related variables in a first model. In a second model, we added education, our labor market variables and the period dummy variables. The total pooled linear effects models reported in columns 1-2 (men) and 5-6 (women) in Table 2 show that there are large gender differences in the association between the family situation of individuals and their chances of having authority. Women who have (had) children and/or are married/cohabiting do not significantly differ in terms of authority from women who do not belong to these categories. However, married and cohabiting men and fathers with children ages 7+ are much more likely to exert authority than are single men and childless men. The sizes of the differences are non-trivial, with married and cohabiting men having a 12.5 percentage point “authority premium” compared with single men, and fathers of children above the age of 6 having a 7-10 percentage point “authority premium” compared to childless men. It is also interesting to note that the intercepts, which in these models indicate the probability that single childless individuals have subordinates, show that the gender gap is very small, even reversed in fact, with women having a slightly higher chance of possessing workplace authority than men (18.1 vs. 15.8 percent). When we add controls for labor market conditions, and education (­columns 2 and 6), the gross fatherhood premiums attenuate toward zero. Additional analyses ­(not shown) indicate that differences in education, work experience and, to some extent, seniority, are primarily responsible for this premium. That is, part of the reason why fathers, and primarily fathers of older children, more often exert workplace authority is that they score higher on education and work attachment. Moreover, the marriage/cohabitation premium for men is also attenuated after the controls are introduced, indicating that more work attachment and higher education among married and cohabiting men seems to be a mechanism partly generating this association.

Extended analyses (available on request) indicate that the “fatherhood premium” is stable and present in all cross-sections. There is also a tendency of a declining male marriage/cohabitation premium over time, but it is present in all cross-sections. For women, the corresponding effects of having (had) children and being married/cohabiting are exclusively insignificantly different from zero in all cross-sections. In other words, the association between parenthood and workplace authority has hardly changed during the time period under study, for men or women.

Results from the fixed-effects model, reported in columns 3-4 for men and 7-8 for women, are partly parallel to the findings from the total models, but also deviate from these findings in important ways, which suggests that the total model effects are biased due to unobserved individual heterogeneity. Similar to what was found in the total model, it seems that fathers gain authority compared to non-fathers. The estimated gains of fatherhood are even higher compared to the estimates from the total model, and the effect appears when children are small (0-6 years old). Moreover, contrary to what was found in the total model, motherhood is accompanied by an increase in authority, but only of a lagged kind: It kicks in when the children are above the age of 20. When we add the controls, the motherhood effect becomes statistically insignificant, and the fatherhood effects tend to decrease substantially, but stay statistically significantly above zero (as does the mother-father difference), implying that much of the fatherhood effect, and all of the motherhood effect, can be attributed to higher labor market-related inputs.

We further find that in the fixed-effects model the marriage/cohabitation effect for women is positive and very close to that of men. There is, in other words, no clear male marriage/cohabitation premium visible in the fixed-effects model, just a gender-neutral positive effect of marriage/cohabitation. For the married/cohabiting coefficients, we can note that the effect for men is more than halved in the fixed effects model without controls. Thus, after individual time-constant heterogeneity is accounted for, the “pure” positive effect on authority for men of entering into marriage/cohabitation is much smaller. This pattern suggests that there is a positive selection of men into marriage and cohabitation. All of the male (and female) marriage/cohabitation premium is captured by the controls, which suggests that marriage and cohabitation for men and women are accompanied by higher investments relevant to the exertion of authority. Because men, but not women, are positively selected into marriage/cohabitation, we only observe a male marriage/cohabitation premium in the cross-sections.

With the caveat that the estimated effects pertain to a select group of men and women (those who exhibit variation in the variables of interest), the results indicate that when men and women become parents, men gain authority, whereas women's chances stay constant. Further, men, who are married or cohabiting are more likely to exert authority, but this seems to be an effect of a selection of authority-prone men into marriage/cohabitation. Taking individual selection and controls into account, the net effect of marriage/cohabitation is not statistically significantly different from zero, for men or women.¹⁰

Conclusion

Research on workplace authority has uniformly shown that women are underrepresented in such positions. In this research we raise two questions regarding the gender gap. First, how has it changed over time? Second, to what extent can the gender gap in authority be attributed to gender-specific consequences of family formation? We extend the second question by asking whether family conditions are causally linked to workplace authority. Because almost all previous studies in the field are based on non-repeated, cross-sectional data, our present knowledge about these issues is limited.

We use individual panel data from the Swedish Level of Living Survey, covering the period 1968-2000, and find that the relative chance for women to reach a position of authority in the labor market improved during the period. The most rapid change appeared in the 1970s, whereas the equalization has moved at a slower pace since the early 1980s. This development does not seem to be unique for Sweden. During the 1990s there has been a stall in women's access to managerial positions also in the United States (Cohen et al. 2009). While this warrants further investigation, one more general cause could be the influx of women into the labor market in the 1970s, with its associated change in attitudes toward women as career oriented and potential managers in both countries, changed (increased) the chances of women attaining authority. After this phase passed and its potential effects were exhausted, however, women's chances to reach authority positions seem to have entered a new steady state, above those in the preceding period, but still significantly below those of men.

We also find clear-cut, gender-specific age profiles in the chance of having workplace authority. As men get older, their chance of holding an authority position increases sharply, peaking in the age span 40-50 years and gradually decreasing thereafter. Women exhibit a different pattern. In their 20s, their chances of exerting authority increase parallel to those of men. Around the age of 30, however, their “career curve” becomes horizontal, and their promotion chances remain significantly below that of men for the remainder of the career.

Our results point out family formation as a watershed generating this gap. As long as male and female employees are single and childless, there is no male advantage in workplace authority. In this group, women are even overrepresented in positions of authority.¹¹ The gender gap in workplace authority appears after parenthood, and is due to the fathers' gain in chances for promotion compared to when they were childless. This is not a selection effect in disguise, as our estimate of fatherhood is within-individual, where all between-individual (time-invariant) heterogeneity has been accounted for. We also found a “male marriage premium” on authority, but this could be explained by individual unobserved heterogeneity, suggesting it is generated by the fact that authority-prone men are selected into marriage and cohabitation.

It can be noted that we find no sign of a “motherhood penalty,” that women's careers suffer when they become mothers. This is somewhat inconsistent with findings for other labor market outcomes, such as hiring (e.g., Correll et al. 2007) and wages (e.g., Harkness and Waldfogel 2003). Whether this pattern is Sweden-specific or authority-specific is hard to tell. Comparatively, the motherhood wage penalty appears to be very small in Sweden (Harkness and Waldfogel 2003), so there might be something about Swedish institutions such as family policies that reduce motherhood penalties across different career outcomes. On the other hand, American studies also show that women's chances to reach managerial positions are unaffected by parenthood (Hopcroft 1996; Huffman and Cohen 2004; Mitra 2003), so motherhood might actually be reconcilable with the possession of authority, across institutional settings.

The differences between the analyses based on the cross-sectional and panel data, respectively, show that the association between family conditions and the gender gap in workplace authority is more complex than that grasped by cross-sectional data. The evidence we have had access to, so far, has relied on such data. Hence we would like to underscore the importance of panel data for future studies in this field of research. Previous estimates of gender-specific effects of family factors on workplace authority are likely to be biased, because men and women are not randomly selected into cohabitation, marriage and parenthood.

While our study constitutes an improvement over earlier analyses in the field, there are some obvious limitations. Our measure of authority captures direct supervisory authority, but ignores other aspects of authority, such as decision-making authority, sanctioning authority and formal hierarchical location. Although a different measure of authority might have yielded a slightly different scenario, we find it hard to believe that the strong effects we find from family-related changes are irrelevant to the attainment of other kinds of authority. Given the interconnectedness of different aspects of authority, it is plausible that family factors also matter for the attainment of these. One consequence of our operationalization is, however, that many, if not most, of those we define as having workplace authority occupy low-level supervisory positions, and people in these positions may not have much influence on important decisions within organizations. In higher-level authority positions, the gender gap is wider, and the process by which men are sorted into these might be different from that operating at lower authority levels. Still, sensitivity analyses that were conducted do not indicate that the influence of family factors is any smaller for the attainment process of higher-level authority.

In addition, family-related changes are not exogenous to the attainment of workplace authority. Hence, we cannot rule out the possibility that individuals may change their family situations as a consequence of whether they occupy a position of authority rather than the other way around. Of course, ordering these kinds of events causally may be a very complicated exercise, aptly illustrated by one of anthropologist Nicholas Townsend's (2002:32) male informants: “And we started discussing about having children and at the same time we discussed that I needed to start preparing myself for a promotion.” As pointed out by Townsend, a strong(er) paid work attachment is part of the culturally dominant “package deal” of being a father, and the time-order of a family change in relation to a work-related change may be hard to disentangle.

Are the results presented here typical for Sweden and other welfare states of a dual-earner type? Because family policies such as generous earnings-related parental leave and extensive public childcare are explicitly designed to facilitate high labor force participation and work-family balance for both parents, our estimates of the effects of family formation, in an international perspective, may be viewed as conservative. However, the generous family policies in Sweden may also provide incentives that – quite contrary to the intentions of policy-makers – exacerbate the influence of motherhood on women's career development (cf., Hakim 2000; Mandel and Semyonov 2006; Rosenfeld et al. 1998; Wright et al. 1995). Swedish women make extensive use of rights such as parental leave and reduced work hours when their children are in their pre-school years. As a consequence, they may thereby sort themselves out of the pool of candidates for authority positions often at the very point in life when promotion opportunities peak. Thus, the dual-earner model, exemplified by Sweden, with its extensive and paid parental leave and the statutory right for parents to work parttime, may promote mothers' labor force participation but, simultaneously, restrict their career chances. In the United States, parents rely on private solutions rather than parental leave programs and public childcare, and women are expected to work more or less continuously and not reduce their work hours after entering parenthood. Consequently, mothers in this institutional context may have better chances to reach managerial positions (Mandel and Semyonov 2006). On the other hand, the dual-earner solution of heavily subsidized childcare, which was gradually expanded during the observation period and now is available to all Swedish parents, should reasonably help career-oriented women. With a one-country design it is neither possible to identify the effects of different kinds of family policies, nor to separate these from other time-varying macro-level factors; that is why further comparative research on the role of family policies for women's chances to reach influential positions is called for.

Leaving this problem aside, the main message of our study is that in the Swedish institutional context, men have better promotion chances when they become fathers, whereas women's chances remain unaffected when they become mothers. At the employee/employer level, we see two plausible, and possibly complementary, explanations of this empirical fact. First, it may be generated by a typically male response to the life event of becoming a father. Men may react to this event in a gender-stereotypical way, and take the traditional household provider role. Entering into an authority position in the workplace is one way of doing this. Women, instead, take the traditional role of caring for their children, which can be combined with an active career involving authority and the like, but only with great difficulty. Thus, the result is that women's careers stall when they become mothers.

Second, employers may begin to treat men differently when they (are about to) become fathers, and interpret the event of a man becoming a father as a signal that he is now prepared to take on more responsibilities at work, an interpretation that is unlikely when a woman becomes (or is expected to become) a mother, due to her anticipated main responsibility for the children (see Bekkengen 2002; Bielby and Baron 1986; Correll et al. 2007; Kugelberg 2006).

Cultural images of being a parent are normative in the sense that shared beliefs and norms of what could be expected of men and women in this life stage constrain their own options as well as what society – including employers – expects of them (cf., Townsend 2002). These behaviors and expectations may reinforce each other, implying that it is far from straightforward to separate supply-side factors from demand-side factors in the reproduction of these gender differences and their labor market-related consequences. In other words, our data do not permit us to adjudicate between these explanations. We consider them equally plausible given the evidence at hand, and additional research is needed to explore their possible interdependencies and relative significance in explaining how family conditions translate into a gender gap in workplace authority.

Notes

References