Work-Family Conflict : The Effects of Religious Context on Married Women ’ s Participation in the Labor Force

Past work shows religion’s effect on women’s career decisions, particularly when these decisions involve work-family conflict. This study argues that the religious context of a geographic area also influences women’s solutions to work-family conflict through more or less pervasive normative expectations within the community regarding women’s roles and responsibilities to the family. We use the American Community Survey linked with community-level religious proportions to test the relationship between religious contexts and women’s participation in the labor force in the contiguous United States–2054 census geographic areas. Using spatial analysis, we find that community religious concentration is related to the proportion of women who choose not to work. Communities with a higher proportion of the population belonging to conservative religious traditions also have a greater proportion of married women choosing not to work outside the home.


Introduction
Women today are faced with increasing demands from both family and work, leading to high levels of work-family conflict for many [1][2][3][4].Faced with these demands, many women are decreasing the amount they work or merely exiting the labor force altogether [5,6].The strategies women employ OPEN ACCESS when faced with these work-family conflicts are not chosen in isolation, but are embedded within larger cultural frameworks [1,7].This paper will further explore married women's employment selection with a specific interest in predicting who will choose to exit the labor force, and what role religion, as a type of culture, plays in this decision.
Religion plays a key role in shaping gender roles as well as discourse about ideal forms of the family, including specific roles and responsibilities for men and women in the family [8][9][10][11].While religion is influential at the individual level, this paper argues that it also plays a role in shaping work-family decisions at the community level.Religion influences one's conformity to social norms of a given area [12], creating a type of moral community.Thus, the religious context of a community will create social norms around the family and gender roles, having an effect on individual's behaviors even if those individuals themselves are not religious.
Whether women choose to remain in the labor force or leave the labor force will be related to their perception of their roles and responsibilities within the family.These views of family roles and responsibilities, what the "ideal" family looks like, are influenced not only by their own religious beliefs, but also by the community context in which they live, which could be more or less religiously dense in nature.As such, we argue that a community's religious context will be related to how many women with families are or are not in the labor force.The paper will examine the religious context using spatial regressions that allow us to control for geographically linked relationships related to our outcome, whether or not married women participate in the labor force.

Women and Work/Family Conflict
The balance between work and family is becoming harder for women today due to increasing demands both at work [1] as well as in the home [2][3][4].While the number of women working is increasing [13], the inflexibility in work hours makes it harder for women to have sufficient time and energy for both work and family [14].This, coupled with the fact that mothers often continue to suffer a "wage penalty" or blocked mobility in the labor market [15], makes work conditions more difficult than ever for women with families today.Studies show that within the household, women still carry out the majority of housework and childcare [13,16].Studies have also shown that work-family conflict increases when there are young children present in the home [17].When faced with these growing demands from both work and family, women often choose to lessen their employment workload [5] and some choose to exit the workforce entirely [6].When faced with these factors, women must enact a strategy for handling the work-family conflict.
Prior research shows that work-family strategies employed by women are embedded within larger cultural frameworks [1,7].This means that the "solution strategies" women have to choose from are constrained by their values and beliefs which are a part of their cultural orientation or framework, all of which are necessarily tied to localized social structures or network connections [18].In her study on female career executives, Blair-Loy [1] found two dominant cultural schemas women are associated with that, in turn, influence the decisions they make about their family and careers.The first is the "family devotion schema" where motherhood and marriage are a woman's primary vocation.The second schema is the "work devotion schema" where the career is seen as a calling or vocation that gives a woman her meaning and purpose, deserving all of her attention.The way in which these female career executives handle their work-family conflict is based upon which was their primary schema.While studies such as this have shown cultural frameworks play a part in influencing women's decisions around work and family, few studies have looked at the specific role of religion.

Family and Religion
Religion is an important influence on a person's cultural framework.Research has demonstrated that a person's religious orientation has an influence on various social outcomes, such as educational attainment and income [19,20].In fact, researchers argue that one of the strongest influences on family life is religion [9,10] and that traditional ideals about family life come from religion.Studies have shown the influence of religion on many areas of family life, such as divorce rates, fertility, parenting, labor force participation, and household labor [21][22][23][24][25]. Thus, when a woman is faced with decisions involving her family, such as work-family conflict, we should expect religion to influence her decisions.
Research shows that religious discourse idealizes certain forms and functions of the family, defining them as legitimate, valuable and morally correct [8,9].This religious discourse creates a framework for women to make sense of and understand gender roles, ideal family arrangements and responsibilities within the family, and is well documented [9][10][11]23,24,[26][27][28][29][30][31][32][33].These gender roles communicate normative expectations that include an understanding of the specific roles and responsibilities for both the man and the woman within the family [10,11].
In particular, research shows that conservative religions tend to promote ideologies that understand the nature of men and women as ontologically unique, each with a sphere of action associated with it.Men's activities focus on the public realm (work) and women's activities focus on the private realm (the home) [8,9,11].More specifically, conservative Protestants as well as most Catholic parishes support various forms of traditional gender roles and endorse a traditional understanding of gender [34][35][36][37][38].
In sum, a woman's perceived "options" for handling work-family conflict are constrained by her own cultural orientation or underlying cognitive schema, which is tied to her localized network or normative connections.Religion plays a strong role in shaping this cultural orientation through the communication of gender roles and understandings of the family that are often an embedded dimension of religious contexts.In a move away from a merely individualistic interpretation of the relationship between religion and family decision making, we argue that just as individuals' decisions regarding work and family depend on the normative community of religion, an entire community can be more or less dominated by these religious normative networks.This will tend to create a community-level sentiment that affects the whole area, not only the individual.

Religious Context
Some schemas and values are domain specific [39] and some cultural schemas are more pervasive over different domains in an individual's life than others [40], but all values are dependent on some social structure or network [41].Likewise, the more pervasive a given network is within a bounded area, such as conservative Protestants for example, the more we would expect correlated values to also be more pervasive within that community thereby influencing all residents' decisions.Specific religious communities should always be considered in light of the wider social context as they are not social islands [42], and the inverse is also true: religious communities will influence the wider social context to increasing degrees with increasing concentrations of religious communities.As mentioned above, religious institutions tend to be highly concerned with the production of family life [9,10], often idealizing certain family forms and functions, claiming them to be moral, legitimate and valuable [8,9,43].As such, the construction of the "moral family" is communicated by religious discourse [9][10][11].While notions of "morally acceptable" families are present in religious communities and individual families through the internalization of religious beliefs, these views regarding which family forms are ideal or "good" can also be influential in the wider community.
Durkheim touches on this phenomenon when defining religion as a "unified system of beliefs and practices relative to sacred things" that unites adherents into a "moral community" ( [44], p. 44).For Durkheim, moral communities are characterized by high consensus regarding communal norms and the religious beliefs that legitimate these norms.Stark [12] argues that religion plays a crucial role in influencing conformity to social norms, and that religion must be considered not only an individual trait but also a property of the community more generally.The more unified a particular context is with respect to religion, the more it can function as a moral community and effectively enforce norms.Thus, the religious context of an area, particularly values relating to family and gender roles within the family, will influence all individuals in that area regardless of personal religious commitments through the creation of communal normative expectations.
In more recent research, the concept of moral communities has helped us understand how social contexts influence human behavior, particularly when looking at deviance and crime [45][46][47], but also more family-related topics such as cohabitation [48].Gault-Sherman and Draper [48] built upon prior individual-level research [49] showing that regardless of one's own beliefs regarding cohabitation, the religious beliefs of their parents had an impact on their own decisions regarding cohabitation.They find that counties with a higher moral consensus regarding the immorality of cohabitation, as measured by a higher concentration of different religious affiliations akin to the present study, do indeed have lower aggregate rates of cohabitation.Thus, for women deciding how to handle work-family conflict, the larger moral community in which they are embedded will influence their decisions.The religious context of an area will create normative expectations about women's role and responsibility in the family which will influence their enacted solution for work-family conflict.Furthermore, in the case of a highly religious context, especially conservative religious communities, women will tend to exit the labor force.

Hypothesis 1a:
The religious context of an area will have an effect on the proportion of women who have families and are not in the labor force.Hypothesis 1b: Areas with higher concentrations of conservative religious groups (such as Catholics and evangelicals) will have more women with families who are not in the labor force.

Data and Methods
The data for this study was taken from two main sources.First, the five-year American Community Survey (2006-2010) (ACS) personal file was used to measure the proportion of married women who work in all communities in the United States.The key advantage of using the ACS is that it is the best available estimate of married working women that includes the entire United States and a way to geographically arrange this data.We restricted our sample to currently married women between the ages of 18 and 65 years.Individual level data from the personal file (age, education, children, sex, race and income) was then aggregated to the area-level.
The area measures employed are the public use micro-area (PUMA) measures from the ACS.We opted to use the PUMA measures as opposed to other options, such as counties, for a few reasons.First, while it is true that PUMA measures are not necessarily related to any other spatial measure directly apart from state lines since each state draws their own PUMAs, there is no direct reason the present model requires county-based measures.Second, when included in models that have no overt need for county-based spatial areas, PUMAs are helpful to some degree because they are a standardized population measure for the entire country.In other words, we do not need to worry about very sparsely populated counties being included in the same analysis as very densely populated counties because PUMAs are 100,000 person area measures.So, while in highly populated areas there will be more PUMAs than in less populated areas, we have gained the ability to standardize our spatial measures by persons within the area and also have the ability to control for PUMAs surrounding the focal area.Finally, PUMAs are more helpful for our analysis than county measures because the data are supplied in PUMA areas and we do not lose specificity by attempting to convert them into counties and end up with areas that are neither fully PUMA measures nor county measures.
The second data source used is from InfoGroup's 2009 Religion Reports, available through The Association of Religion Data Archives (ARDA).ARDA staff built their categorization scheme based upon Steensland et al. [49] and used Melton [50] and Mead and Hill [51] to categorize additional denominations not found within Steensland et al.'s categorization.InfoGroup subsequently utilized the resulting categorization when organizing the data [52].Following Steensland et al. [49], we collapsed various religious traditions with lower adherence rates (e.g., Hindus, Buddhists, etc.) into an "other" category with the exception of Mormons who also have highly conservative family values [53].This data includes the number of members affiliated with each religious tradition, which was then used to calculated area religious proportions using the 2007 census population estimates.This allowed us to look at the religious context of the area through the proportion of different religious groups present in the PUMA of interest as well as in the surrounding PUMAS.

Analytic Method
We argue here that the religious proportion of any given geographic location will be associated with different proportions of working married women within those areas and have thus opted to use spatial models in the present study.An alternative approach could have employed multi-level models to predict whether the religious concentration of a given area was related to any given individual married woman's choice to work or not work.We opted for the present spatial models for two reasons.First, our argument follows Durkheim, Stark and Gault-Sherman and Draper's work outlined above on moral communities arguing that religious proportions are not only an individual trait but also a communal trait.We additionally treat the proportion of married women within different communities who work outside the home as a trait of that community.Since this treats both the community's religiosity as well as the tendency for married women to work or not work as community traits, our dependent variable (whether women work or not) is not on the individual level.Second, moral communities should not be thought of in isolation from one another.This means that one community can affect another community, and so proximity is important.This is exactly what the present spatial models do.
The spatial effects of religion on the proportion of married women working outside the home are estimated with spatial lag models, which insert a spatial autoregressive parameter into the regression equation [54].The spatial lag term was used, as opposed to accounting for spatial relationships within the error term of the regression equation, after following Anselin's [55] recommended diagnostic tests.There are two ways to think of spatial relationships [54], only one of which implies primarily the spatial dependence of the dependent variable over multiple communities.We are interested in controlling for the spatial dependence of our outcome in this study proportion of working married women, because the proportion of working married women has a non-random geographic distribution in the United States, as indicated by the robust Lagrange Multiplier in Anselin's [55] diagnostic tests.This means that the proportion of working married women in nearby communities potentially influences whether or not mothers in a local community work outside the home-a diffusion effectleading to regional differences and pockets of "traditional" or "egalitarian" family values.This implies a longitudinal trend creating a non-random geographic pattern that we, using cross-sectional data, are merely interested in controlling for [56].Thus, when we look at the proportion of married women who are working, the spatial lag term helps us avoid falsely attributing explained variance to our key measures when it should instead be attributed to the proportion of working women in surrounding communities.This takes into account the idea that geographic units are not free of outside influences, but rather are part of a larger geographic area where there is interdependence and a transferring of ideas and people between the areas [57].

Measures
The key dependent variable in this study is the proportion of married working women in each spatial area, so only those who were currently married and between the ages of 18 and 65 were included (See Table 1).The family employment status variable (FES) was used to determine which women were working.Labor force was determined by using the employment status variable in the census data.This variable identifies women who worked at any time during the previous week, women who were available for work but were temporarily laid off, women who had jobs but were off (not laid off) during the previous week, and women who did not have a job in the previous week but were actively looking for a job.Thus, this variable captures those who are a part of the labor force, including those who may be unemployed but are searching for a job.We choose this variable because it best measures not only those who were able to get a job, but rather those whose preference was to be in the labor force-those women "choosing" to work.Those who responded as either "married couple family: husband and wife in the labor force" or "married-couple family: husband not in the labor force, wife in the labor force" were coded as a '1' reflecting married women who were working, while all others were coded as a '0'.Once this dummy variable was created, the measure was aggregated to the area level, producing a measure reflecting the proportion of women in that area who were married and working.Because it is a proportion, it is a continuous measure, with a range of 0-1, where 0 is no married working women in the area and 1 reflects that 100% of the population of married women is working.Additionally, we included an adjusted wages measure that is naturally log-transformed and then aggregated.By taking the log of the adjusted wage we were able to control for the skewedness of the wage distribution in each area before aggregating to the higher level.We also control for how educated the area is, the proportion of white residents, the proportion of females, the proportion who have children still living at home, and the average age of the area.
We have included the proportion of the population in each area that is a member of each different religious tradition.This allows us to measure and control for the effect that greater or lesser presence of each religious tradition may have on the proportion of married working women of that area.Related to this, we have also included a measure similar to the Herfindahl index, which essentially is a measure of dissimilarity [48].Higher scores reflect an area where any single resident will tend to be religiously heterogeneous from other residents in the same area.
Finally, we have created a spatial lag term for our analysis.The term relates specifically to the dependent variable, the proportion of married working women, allowing us to control for the effect that variations of the dependent variable in adjacent areas may have on our focal area's outcome.The spatial weight is a weighted average of the endogenous variable in all contiguous geographic locations to the focal area [58][59][60].The connectivity distribution (available upon request) shows that there are no spatial islands that we need to be concerned about and the distribution is relatively normal.

Results
In looking at the boxmap (Figure 1), we find that while there is great variability in how many women are working outside of the home, there does not appear to be a regional pattern.Because there does not appear to be a regional geographic pattern in the United States where more or less married women are working as compared to other regions, we did not test for a spatial regime.This means it is possible that the mechanism behind women working functions in relatively the same way in all areas of the United States.In order to discuss how the religious concentration of one area affects those around it, we have included local spatial autocorrelation maps (LISA maps) which graphically show areas where there is positive spatial autocorrelation (see Figure 2) [55].Positive spatial autocorrelation is defined by areas where a statistically significant relationship is present between a measure and its spatially weighted counterpart.For example, if the proportion of married working women is high in a focal area and the first-order (i.e., contiguous neighbors) spatially weighted married working women term is also high, then the spatial autocorrelation is positive and significant.If the proportion of married working women is low and the first-order spatially weighted term is low, then the spatial autocorrelation is negative and significant.These are the two primary relationships we are interested in, but there can also be low-high area relationships as well as high-low relationships.This relationship is tested using the Moran's I statistic [55], allowing for us to determine if the spatial autocorrelation is significant or not.This test yields a significant Moran statistic of 0.2646 (p < 0.001).From these results, it appears as though there is some type of pattern between the proportion of married women working in an area and the area surrounding it.
Table 2 shows the regression results.Higher average levels of education and a greater number of women in general within the focal area both have a positive association with the proportion of married women who participate in the labor force.The proportion of families with children is also positively related with married working women, which could indicate a generation effect where younger women are more likely to work because areas with a higher average age also have a negative association with married working women.The proportion of women who are white also has a negative relationship with the proportion of married women in the labor force.In support of hypothesis 1, the religious context of an area does have an association with the proportion of women in that area who participate Lower Outlier (41) in the labor force.We find a positive relationship between areas with a larger proportion of mainline Protestants and the number of married women in the labor force and, conversely, that areas with a larger proportion of Jewish or Mormon membership have fewer women participating in the labor force.In support of hypothesis 1b, we found that the proportion of evangelicals in an area has a negative association with the proportion of women in that area who participate in the labor force, although the relationship between the Catholic proportion and proportion of working women was not significant.

Conclusions
We have argued that religion is influential not only for individual women's attitudes regarding gender roles and subsequent employment behavior, but that the religious context in which they live is also important.Norms guiding behavior are always tied back to a social structural source.With this in mind, the link between religion and family values and normative expectations should not merely be thought of as an individual-level measure, but also a communal-level measure as those religious networks are more or less pervasive within a given community.While prior research has shown that the personal views are influential [27,32,61,62], we have focused on the contextual influence here.We find that areas with a greater proportion of religious groups that emphasize traditional gender roles are negatively related to the number of married women who participate in the labor force.On the other hand, areas with a greater proportion of religious groups emphasizing more egalitarian gender role ideologies, such as mainline Protestants, tend to have more married women participating in the labor force.As such, the religious context of an area is an important factor to be considered when considering how women handle work-family conflict.
The present analysis has some data limitations that should be kept in mind.First, we were unable to include the effect of personal, individual-level religious beliefs as this information is not available in the ACS.As we discussed above, the benefits of using the ACS are such that the present study is a valuable step in the right direction of taking into account the religious context in which one lives.This does mean that the present work cannot show that a moral community is more important than individual-level religious beliefs, but the present work does give good reason to take one's communal context seriously in future work.Future research should attempt to use data including both individual-level and communal-level religion measures that lends itself to multi-level modeling, even if doing so limits the analysis to less than a national sample.Secondly, the present analysis is cross-sectional and is unable to directly test for causal relationships.It is possible that one's employment status or activities in life actually contributes to or detracts from an area's religious membership and an individual's propensity to be a member of a congregation.While this is a possibility, it seems more theoretically plausible that the relationship would be as presented in this study-religious beliefs inform views regarding gender roles.Accordingly, these normative expectations are more or less pervasive in a community, the weight of which influences all within an area and relates to participation in or abstention from the workforce.

Figure 1 .
Figure 1.Boxmap of the distribution of married working women.

Table 2 .
Spatial regression of proportion married women working.