Gender (in)equality within the household and business start-up among mothers

The predominant perspective in the research on women’s entrepreneurship considers gender inequality within households as a persistent feature of most institutional contexts. This feature persists because informal institutions are subject to strong inertia (North 1990). The central premise is that even though the fast-changing world outside the home has moved towards greater gender equality, the roles of men and women inside the home have changed relatively little (Calasanti and Bailey 1991; Skolnick 2009). Throughout the world, women continue to perform “the lion’s share of unpaid caring work” (Ahl and Nelson 2015:3), and domestic work and obligations, including child care, “fall disproportionately to women, even if women work equal or longer hours than their male spouses” (Baughn et al. 2006: 689).

The above observations are supported by sociological studies showing that when men and women become parents, they tend to resort to more traditional gender roles within the household (Dribe and Stanfors 2009). Lappegard (2008: 158) study of parental leave in the Nordic countries indicates that “since the common parental leave period is today considered more of a mother’s right, fathers are faced with more obstacles to taking more leave”. Additional research from Sweden shows that a father might not use a large share of parental leave because “he has to weigh the possible negative reactions from workmates, supervisors, and the employer against the amount of parental leave he wishes to take” (Sundström and Duvander 2002: 443).

Thus, in making career choices, mothers of young children need to take into account how to best balance their domestic and economic obligations (Jennings and McDougald 2007). In institutional contexts with formal arrangements that support work-family balance, having wage employment provides mothers with the best conditions for simultaneously managing such domestic and economic obligations. Indeed, the benefits of arrangements such as paid parental leave are connected to wage employment and cannot be entirely realized by mothers starting their own businesses.

Under these premises, in institutional contexts with a formal family policy system, women are pushed to start businesses mainly because they are unemployed or because they face unfavorable conditions in the labor market (Thébaud 2015), such as those experienced by immigrants who encounter more labor market obstacles than nonimmigrants due to a lack of language skills and local networks (Evans 1989). This perspective in the entrepreneurship literature is also supported by the sociological literature on the so-called welfare state paradox (Mandel and Semyonov 2006). This literature holds that by allowing reduced working hours and long parental leaves, extensive family policies contribute to preserving gender inequality within families, as they allow women to maintain their traditional role as primary nurturer and caregiver in the family. Moreover, it is argued that extensive family policies might even increase employers’ discrimination against mothers, making it more difficult for mothers to find wage employment because such policies render “mothers less attractive to employers who prefer employees with an uninterrupted commitment to work” (Mun and Jung 2018: 511).

In summary, according to the mainstream view detailed above, mothers of young children in Sweden turn to entrepreneurship mainly as an economic necessity—that is, in response to unemployment and job-loss (Hughes 2003) or because they are in a position in which finding employment with full job benefits may be difficult, such as for immigrants. Thus, being unemployed and/or being an immigrant are expected to be the factors most strongly associated with entrepreneurship among mothers.

Another less visible perspective does not consider gender inequality within families as a persistent feature of most institutional contexts. Building on the assumption that institutions can change, the essence of this perspective is that formal institutional arrangements—such as extensive family-oriented policies—may contribute (at least in part) to changing informal institutions, such as traditional expectations regarding the division of responsibilities and roles between men and women within the household (Neergaard and Thrane 2011). Specifically, as supportive family policies have increased women’s participation in the labor market (Schultz 1990), they have also enhanced women’s independence and contributed to challenging traditional gender inequality within the household (Craig and Mullan 2011; Haas 1992; Hobson 1990; Hook 2010). In more detail, the sociological literature—e.g., gender and welfare state theories—holds that the national context and in particular welfare policies have influenced normative gender expectations about behavior, especially within the household (Hook 2010). McDonald (2000) describes this process as the ‘gender revolution’. During the first part of this process, women enter the public sphere of education and employment with the help of policies (e.g., parental leave) that make women’s employment and a dual breadwinner system the norm. During the second part of the process, men enter the private sphere and share “responsibility for the care of children equally with their female partners (just as women increasingly share the provider role)” (Bernhardt et al. 2008: 278).

Research has, for example, shown that “men who earn approximately the same income as their spouses perform more housework than men who are primary breadwinners” (Thébaud 2010: 332) and has discussed the “increasing convergence of sex roles” (Coverman and Sheley 1986). This type of household resembles what Giddens (2005) calls the “democratic family”, in which there are no fixed norms regarding who should do what. Key family issues are subject to negotiation and joint decision-making between the parents (Ahlberg et al. 2008), and normative expectations are characterized by “double emancipation”—women can be workers as well as carers, and men can be carers as well as workers (Klinth 2002a, b). The dual-breadwinner/dual-career model of countries such as Sweden is reflected in a new image of fatherhood (Dribe and Stanfors 2009; Jensen 2000). Paternity leave becomes particularly important because “it makes fathers available for time-inflexible housework and childcare, enabling mothers to return to the workforce sooner or invest in their career” (Patnaik 2015: 9). Research also shows that men who report bearing more responsibility and performing more hours of childcare at home also reported taking more parental leave (Haas and Hwang 2008).

Under these premises, in institutional contexts with an extensive family policy system, such as Sweden, mothers can be expected to make career choices by negotiating with their partners who will make use of the parental benefits offered by the government (e.g., paid parental leave) and take responsibility for childcare. Hence, entrepreneurship is not necessarily a fallback strategy but may reflect the pursuit of an opportunity facilitated by the egalitarian trend within the family context (Brush and Hisrich 1991). Baughn et al. (2006: 692) discuss pull factors contributing to entrepreneurship by mothers, suggesting that “entrepreneurship may be more attractive relative to wage/salary employment in societies characterized by higher levels of gender equality”.

In summary, according to this perspective, the amount of paternity leave taken by fathers is expected to be the factor most strongly associated with entrepreneurship by mothers in Sweden.

Women’s business ownership in Sweden has always been a reflection of formal, regulative institutions, and market conditions. Before 1846, women could not own a business. At that time, widows and divorced or unmarried women were allowed to engage in petty trade to support themselves. Married women could control their own income in 1874, and single women became major at age 21 in 1884, but it took until 1921 before married women became major and women could vote. However, women’s business ownership then increased rapidly. Women’s share of business ownership was 27% in the 1930s and 1940s (Lundström 1996). In Sweden, the mom-and-pop store was a mom store—women had milk, bread, and delicatessen shops, laundries, and mangling services. The 1950s and 1960s saw the rapid rationalization of retail, and many small shops were closed. Female labor was instead absorbed by the rapidly expanding manufacturing industry but even more so by the growing public sector in the late 1960s and 1970s. Small business ownership declined overall, and women’s ownership declined in particular—their share of business ownership decreased to approximately 15% in 1972. However, as small businesses took an upward turn after the oil crisis in 1973, women’s share of businesses rose again, particularly in the growing service sector—by 1990, it was over 30% (Svanlund 2011). The 1990s saw another economic crisis, which was cured by drastic cuts in the public sector and the privatization of formerly government-owned operations in care, health care, and education, accompanied by government policy and programs to stimulate women to start businesses in these areas (Berglund et al. 2018). In 2016, women’s share of business ownership had risen to 38%. Most of these businesses were in personal or cultural services (33%), business services (21%), retail (9%), care (7%), and farming (7%) (SCB 2018).

Women’s participation in the labor force has been, and is, substantial in Sweden. In 2017, the figure for women aged 20–64 was 85%, and it was 89% for men (SCB 2018). The gender pay gap has slowly diminished over the years—the weighted gap, which controls for age, educational background, full-time/part-time, sector, and occupational group, was 5% in 2016, and the unweighted pay gap was 12% (SCB 2018). A dual breadwinner system is both the norm and a necessity.

Women’s high labor market participation was made possible by several policy reforms. An important reform was the introduction of separate taxation in 1974. Before then, it sometimes did not pay for a married woman to work—her income, which was added to her husband’s when doing the tax return, was consumed by a very high marginal tax rate. The family policy also changed. Beginning in the 1970s, public, universal daycare and preschools, as well as world-renowned academic education for preschool teachers, were established. Today, every child aged 1 has the statutory right to subsidized daycare and preschool that is affordable for everyone.

In 1974, paid parental leave that could be divided between the mother and the father as they wished replaced the former system of maternity leave. However, due to entrenched gender roles and resistance from employers, fathers’ uptake of this leave was initially negligible. To stimulate more fathers to take parental leave, 30 days were earmarked for them in 1995, 60 days were earmarked in 2002, and 90 (of a total of 480) days were earmarked in 2016. The idea is that if legislation and economic incentives change, the attitudinal change will follow. In 2017, fathers used 28% of all parental leave days, so the policy seems to have worked, albeit slowly. The 480 days can be allocated as parents wish during the first 8 years of a child’s life, and fathers tend to take parental leave more often when the child is a little older (SCB 2018). Parents also have the statutory right to stay home from work with a sick child, while receiving pay, for a maximum of 60 days per year. Men took 38% of such days in 2017 (SCB 2018).

Time studies show that Swedish cohabiting women perform approximately 55% of unpaid work, including housework, dependent care, maintenance, shopping, and commuting for work, but cohabiting parents of small children spend much more time on unpaid work than others: 39 h per week for women and 32 h per week for men compared to 26 and 21 h per week, respectively, for all parents (SCB 2018).

These circumstances make Sweden the ideal context for exploring the two aforementioned views of the intersection of entrepreneurship and motherhood: the “mainstream” view and an “alternative” view.

In order to conduct the study, we constructed a longitudinal dataset by combining four longitudinal Swedish databases: (1) the Longitudinal Integration Database for Health Insurance and Labor Market Studies (LISA for its initials in Swedish), which provides yearly data on all residents of Sweden, including family relationships; (2) the Jobbregistret (work register) database, which provides yearly data on all Swedish inhabitants’ income sources and work activities; (3) the Labor Market Statistics (known by the acronym RAMS in Swedish) database, which provides yearly data on all firms registered in Sweden; and (4) the multi-generational-database, which provides information on couples (i.e., if they are married or if they are living together and have children together) as well as on biologically linked families (i.e., parents and children). As these databases provided by Statistics Sweden contain annual information about individuals, our sample and analyses are based on annual observations of data. These are official statistics, which are considered highly accurate and reliable (cf. Baù et al. 2017; Wennberg et al. 2011; Wiklund et al. 2013).

To identify a sample of mothers of newly born children, we started with an initial sample of all children born in Sweden between 2000 and 2014. From these data, we identified the mothers—16 to 55 years of age—of each newly born child and collected longitudinal information about their work status (e.g., employed or unemployed), background (e.g., immigrant), parental benefits (e.g., paid parental leave), work experience, education, income, and use of parental benefits. We collected identical information about the women’s partners.

We used a Cox proportional hazard model to test what factors explain the likelihood that mothers will start a business (Cameron and Trivedi 2013; Cleves et al. 2008; Rabe-Hesketh and Skrondal 2008; Royston and Lambert 2011). This approach allows an event-history analysis incorporating longitudinal data with time-varying variables to predict both the timing and the occurrence of an event (Allison 2014); it also analyzes the influence of predictor variables on the occurrence or nonoccurrence and timing of specific events (Morita et al. 1993). Our model considers semiparametric proportional hazard regressions. Thus, the hazard rate is calculated as the product of an arbitrary and unspecified baseline hazard rate h₀(t). The vector includes a number of covariates and regression coefficients to be estimated x_iβ (for details see, e.g., Efron 1977: 229–244).

However, as our analysis is based on annual data, we are only able to observe failures in a discrete manner (‘ties’). Our dependent variable is dummy coded to indicate whether the mother starts a business. Indeed, we only know which and how many mothers become entrepreneurs from year to year; we cannot distinctly order their entry times (i.e. failures) within each period (see Cox and Oakes 1984). Therefore, we adopted the method of Efron (1977) within the Cox proportional hazard framework to correct the partial likelihood function in order to cope with the existence of ties and to obtain unbiased and consistent estimates.

In particular, we identified as a fail event the entrepreneurial entry of a mother of a newly born child during a period of up to 8 years after the child’s birth. We chose 8 years as the maximum period for following each mother because this is the length of time during which parents can make use of paid parental leave in Sweden. In the period 2000–2014, we identified about 1 million mothers of newly born children. Out of this million, 60,000 started a business during the time they (and their child’s fathers) could have made use of parental benefits.

Further, we excluded those mothers who were already engaged in entrepreneurial activities before the birth of the child. Moreover, as we make predictions based on gendered roles within the household (e.g., gender inequality within the household), we selected only those mothers who were living together with their partners in a household. This selection yielded a final sample of more than 700,000 mothers. Out of these, 45,000 mothers started a business in the given period.

In order to test the predictions of the mainstream view, we employed two independent variables in our analysis: the mother’s unemployment and the mother’s immigrant status. Mother’s unemployment was dummy coded as “1” if the mother was unemployed for at least 1 month in the year before the birth of the child, “0” if she was permanently employed. Similarly, as we want to identify women who face labor-market obstacles due to their lack of language skills and local networks, we test for mother’s immigrant status (Evans 1989) coded as “1” when the mother was born in a country other than Sweden, “0” otherwise.

We introduced several control variables at the individual, household, and regional levels that could affect our results.

At the individual level, we accounted for the mother’s age, years of education, prior entrepreneurial experience, parental leave, and the number of offspring. We controlled for mother’s age because it may affect the probability of engaging in entrepreneurial activities (Langowitz and Minniti 2007). Then, as human capital affects the propensity for starting a business, we controlled for mother’s years of education (Ucbasaran et al. 2003), coded as a continuous variable that indicates the number of years a person spent in school and at university. Research has shown that education is an important factor in explaining a person’s decision to start a business, especially for women in high-income countries (Allen et al. 2007). Similarly, it has been shown that prior experiences as an entrepreneur might provide important skills and social capital that could be exploited in a novel initiative (Baù et al. 2017); thus, we controlled for mother’s prior entrepreneurial experience, dummy coded as “1” if the mother had an entrepreneurial activity terminated or sold before the birth of the child, “0” otherwise. Moreover, to provide the complementary perspective of the partner’s parental leave used as an independent variable, we control for the mother’s parental leave, calculated as the mother’s parental benefits paid in the year, divided by the mother’s individual’s income in the same year. Finally, we controlled for the number of offspring in the household, as the number of children might change the family structure by, for example, introducing additional time constraints and parental responsibilities.

At the household level, we included the partner’s unemployment, immigrant status, years of education, prior entrepreneurial experience, and family disposable income. In order to complement the independent variables used for testing the mainstream view within the household, we controlled for partner’s unemployment, dummy coded as “1” if the partner was unemployed for at least 1 month in the year before the birth of the child, “0” if permanently employed. Similarly, we controlled for partner’s immigrant status, coded as “1” when the partner was born in a country other than Sweden, “0” otherwise. Thus, given that the partner’s education might affect the ‘double emancipation’ of both women and men as workers as well as in their careers (Klinth 2002a, b), we controlled for partner’s years of education, which we coded as a continuous variable that indicates the number of years a person spent in school and at university. Moreover, we controlled for partner’s entrepreneurial experience, dummy coded as “1” if the partner is an entrepreneur, “0” otherwise. The variable is lagged 1 year.

Finally, the family business literature has largely investigated the propensity of families to self-finance entrepreneurial initiatives. Indeed, a large household income may generate slack resources that favor entrepreneurial initiatives by individual members (Alsos et al. 2014). Thus, we controlled for the logarithm of the household’s disposable income. The variable is calculated by Statistics Sweden as the sum of the individual disposable incomes of the household members, expressed in hundreds of SEK.¹ The variable is lagged 1 year.

Further, we controlled for the characteristics of the municipality of the household’s residence, which can affect business start-up. Extant research has observed that start-ups per capita are relatively more concentrated in metropolitan areas and least frequent in rural areas (Westlund 2011). Therefore, following the work of Westlund et al. (2014) and Nilsson (2015), we separated Swedish municipalities into two categories and coded municipality type as a dummy variable equal to “1” to indicate an urban context, which includes both metropolitan and urban municipalities that supply a large variety of services to the individuals and firms based in that location. Thus, we coded the rural context as “0”, which includes rural and sparsely populated rural municipalities that host a limited variety of services. Finally, we controlled for time by including the number of years passed since the birth of the child.

In this study, we identified two potential sources of sample selection bias. The first one is induced by having or not having a child within the time frame (2000–2014) chosen for the study. Thus, using Heckman’s (1979) two-step procedure, we identified the mother’s status of being a single child as an instrumental variable for calculating the inverse Mills ratio to include in the analyses. The instrumental variable must be correlated with the first-stage probit model’s outcome (i.e., having a child) but not with the second-stage model’s outcome (i.e., mother’s entrepreneurial entry) (Kennedy 2008; Wooldridge 2002). In detail, we calculated a probit regression per each year of our sampling period, comparing all women who had a child in the given year with those in the age range 16–55 who had no children in the same year.² Therefore, we predicted an inverse Mill’s ratio from each of the year-based probit models that we included in our analysis (i.e., Mill’s no children).

The second potential sample selection bias is induced by the choice to include only mothers living in a household with their partner, thus excluding single and divorced mothers. As we mentioned, this choice was driven by our interest in testing the two views of gender inequality within the household. Nonetheless, this choice has to be addressed by using Heckman’s (1976) two-step procedure. In this case, the instrumental variable used for calculating the first-stage probit model’s outcome was the mother’s divorced parents, dummy coded as “1” in the affirmative case. Indeed, Mustonen et al. (2011) found that in Finland, women and men from divorced families are more often divorced or separated at the age of 32 years than are those from nondivorced families. Moreover, they also found that parental divorce is associated with poorer intimate relationship quality for women. Thus, we use the predicted inverse Mill’s ratio in our analysis (i.e., Mill’s no partner).

We also performed some important additional robustness tests to verify our main results. First, looking at the immigrant background of both the mother and the partner, we considered different alternatives. (1) We distinguished between Scandinavian (i.e., Sweden, Finland, Norway, and Denmark) versus non-Scandinavian backgrounds; (2) we distinguished between European (i.e., the 28 members of the European Union) versus non-European backgrounds; (3) we distinguished between individuals born in the European Union or the United States versus other countries. The results were consistent with our analysis, validating our results in comparing the first group (including Swedish-born mothers) with the control group.

Second, in considering unemployment status, we tested alternative definitions calculated with a 1-year lag. We considered (1) unemployment status in the year before; (2) number of years of continuous unemployment; (3) number of months of unemployment in the year before; (4) number of months of continuous unemployment. Also in this case, the results were significant and consistent with our main analysis.

Third, we conducted an additional robustness test to proxy the status of economic need. In particular, we considered whether the family income was below the poverty line in the previous year. Persistent risk of poverty refers to persons living in a household where the equivalized disposable income is below 60% of the national median during the reference year and during at least two out of the preceding 3 years (Eurostat³). This additional test provided results consistent with the mother’s unemployment status.

Fourth, we also measured household disposable income by considering the individual disposable income of the mother and her partner. The results are consistent and significant.

Finally, during the review process, we recognized the opportunity to test for same-sex couples. We isolated in our dataset approximately 1400 couples in which mothers shared a household with a female partner. In these households, we observe about 50 events of entrepreneurial entry of mothers. Excluding same-sex couples from the dataset, coefficients and significance levels were consistent with our full model.

Our study provides important theoretical and practical implications. First, our study contributes to the research on women’s entrepreneurship and work-family balance. Studies in this tradition hold that mothers are less likely to pursue entrepreneurship “because they lack attractive employment options in contexts in which supportive institutions are in place” (Thébaud 2015: 671). Instead, our study shows that while lack of good employment options plays a role in explaining entrepreneurship by mothers in contexts characterized by supportive work-family policies, the amount of paternity leave taken by fathers is also an important factor. Thus, our study challenges the main view within this literature that “women are expected to be the primary nurturer and caregiver in the family, giving priority to serving the needs of any descendants” […] while “men are expected to be the good economic provider for the families” (Jennings and McDougald 2007: 752). Going forward, our study shows the importance of a contextualized perspective on entrepreneurship. While the current literature on work-family balance within the women’s entrepreneurship literature has begun to compare women’s entrepreneurship across different institutional contexts (Estrin and Mickiewicz 2011; Noseleit 2014; Thébaud 2015), some of the key assumptions underlying this research—such as gendered roles within households—are very a-contextual.

More generally, our study contributes to the institutional perspective on female entrepreneurship by highlighting the interplay between formal and informal institutions within one specific context (Baughn et al. 2006; Estrin and Mickiewicz 2011; Welter et al. 2014). Specifically, our study shows that formal institutions that support work-family balance might have also contributed to altering gendered expectations of roles and responsibilities within households and thus provided specific normative support for women’s entrepreneurship by making it socially acceptable—or even the norm—for fathers to assume the role of the main child carer.

Our study also has important policy implications. Our findings can also be discussed with regard to the start-up propensity of women in general. Our findings suggest that institutional arrangements—such as a well-functioning parental leave system—that have facilitated mothers’ participation in economic activity might also enhance mothers’ opportunities to become entrepreneurs as long as their partners become active childcarers. Thus, while the study’s findings speak in favor of parental leave schemes that are gender-neutral—that is, parental leave benefits that can be transferred between partners—they also show that it is important that such transfers take place within the household and that fathers take paternal leave.

Beyond Sweden, our findings are also relevant for other countries that wish to promote women’s pursuit of opportunity-based entrepreneurship. We flip the traditional focus on barriers to women’s entrepreneurship—and how to alleviate or remove them—to inspire policymakers to consider an enabler of women’s entrepreneurship—i.e., the parental leave taken by fathers. Our approach also underscores more specifically the importance of family policies in general and parental leave systems in particular.

While offering a new perspective on women’s entrepreneurship, our study has limitations that offer opportunities for future research. We used secondary data; therefore, we cannot truly capture gender dynamics within the household. Additional research might use longitudinal qualitative data to better capture such dynamics. Our research design focuses only on mothers who live with their partners in the same household. We have chosen this approach because we were interested in contrasting different views on gender inequality within households. Additional studies should expand our research by also investigating single mothers, divorced couples, and different or extended types of households.

Our study opens up a wide array of research opportunities. We offer the following suggestions. First, our study shows that the participation of the partner in the division of childcare (e.g., the use of parental leave) and employment conditions are important factors for women entering into entrepreneurship. Future research should examine whether the same pattern is observed among men. Second, given that women’s childbearing and child-rearing are often associated with a ‘motherhood penalty’ (Correll et al. 2007), research should explore any career and earning consequences for men actively using the parental leave. For example, would men be entitled to ‘fatherhood bonuses’ or subjected to a ‘fatherhood penalty’? Third, research exploring effective mechanisms contributing to the change in normative expectations regarding the nongendered division of housework and family care and its impact on entrepreneurship activity among both men and women would help further our contextualized understanding of the phenomenon. As such, our research encourages more studies on entrepreneurship to adopt a household perspective (Carter 2011) when exploring the motivation for entering entrepreneurship.