Gendered transitions to self-employment and business ownership: a linked-lives perspective

ABSTRACT We apply the sociological lens of linked lives to show how household contexts channel transitions to self-employment in ways strongly differentiated by gender. We investigate the impact of demographic transitions to marriage, cohabitation and having children on the transition to self-employment using fixed-effects models on 10 waves of the UK’s nationally representative survey, Understanding Society. Men’s transitions to self-employment and separately to business ownership are remarkably impervious to the arrival of a new child in the household. In contrast, second births raise the odds of self-employment for women and have a strong and statistically significant association with business ownership, highlighting the role of birth parity as a household influence. Within the subset of opposite-sex couples, lives are indeed linked: a partner’s long hours precipitate the other partner’s transition into self-employment for men and women. However, the effect is asymmetric to the extent that women are much more likely to have a partner working long hours. Marriage is associated with a much higher likelihood of transitioning to business ownership for both men and women, which does not hold for self-employment overall.


Introduction
The number of self-employed people in the UK has risen considerably in recent years, from 3.3 million (12% of the labour force) in 2001 to 4.9 million (15% of the labour force) in 2019 (ONS 2020).Women comprise a growing segment of the UK's self-employed, in line with developments in Australia, Canada and the USA, although men still account for about twice as large a share (68%) as women (Henley 2017).This growth in self-employment and women's share of it motivate this article's exploration of the gendered pathways into self-employment from employment in the UK.We take inspiration from Jayawarna, et al (2020) who draw on the life course perspective to posit that the lives of individuals become linked within the life course when they: establish a separate household, marry or decide to become partners, they have children, divorce or decide to retire; and each such stage of the life course has specific implications for household relations.For instance, we know that events such as the arrival of children change household membership and allocate additional responsibilities and roles to specific partners prompting partners to rethink their engagement in work in order to keep the household productive and to maintain financial well-being (Hutchison 2011).Taking a life course perspective, we propose that gender differences relating to selfemployment are accentuated or diminished at key turning points in life courses events.Societal institutions such as labour markets and education shape how these transitions occur across different contexts.
In recent years, there has been a dramatic increase in mothers' labour force participation in the UK, reaching 75% in 2019, even higher than the 70% participation of women without dependent children, according to estimates from the Labour Force Survey (ONS 2019).Both parents are employed in about seven in 10 opposite sex couples with children (ONS 2019).Although the male breadwinner pattern is no longer the default, neither is equal participation by men and women, especially those with children.These changes call for renewed attention to household ways of organizing, linkages across generations and specifically for those in couples, how lives are linked.We examine such linkages in relation to self-employment and separately to the sub-category of business ownership.
The conditioning of individual agency on ties with others is a key tenet of the life course perspective (Settersten 2015a), so that a full understanding of an individual's actions and positions involves reference to key others in that person's life, both within and across generations (Elder 1994;Moen 2003; Moen and Sweet 2003).The effects of demographic life course transitions, notably having children, differ substantially by gender.Motherhood is strongly associated with pay penalties (Budig and England 2001;Harkness and Waldfogel 2003;Musick, Bea, and Gonalons-Pons 2020) and the gender wage gap (Bryson et al. 2020).Birth parity also plays a significant role in women's work patterns in the UK (Stewart 2014); more children result in more intense childcare needs, which are associated with reductions in mothers' paid work intensity (Schober 2013), since care remains socially ascribed to women.
Women's diminishing labour market position after having children (Costa-Dias, Joyce, and Parodi 2018) occurs in various ways, for instance, through work exit (Kanji 2011;Jayawarna, Marlow, and Swail 2020) and occupational downgrading linked to moving to part-time employment (Connolly and Gregory 2008).Such studies on gender inequalities tend to focus on employment, to some extent overlooking whether transitions to self-employment are driven by the need to reconcile work and care (Craig, Powell, and Cortis 2012), as well as in contexts in which policy is less supportive of work−life reconciliation (Thébaud 2015).Having the highest formal childcare costs in the OECD (OECD 2022) could, to some extent, place the UK in the category of less supportive of work-life reconciliation.Brush et al (2009) signal the significance of motherhood in women's entrepreneurship by using it as a metaphor for the household/family context which interacts with the macro and meso environments shaping expectations of what are acceptable choices, for example, for mothers.Institutional support for children is an important aspect of this macro-environment: in contexts with less support women may be pushed into self-employment whereas in contexts with more support, both men and women may focus on more entrepreneurial forms of self-employment (Brush, de Bruin, andWelter 2009) Thébaud (2015); Cheraghi et al. (2019).This balance matters, not least because entrepreneurship or new business creation creates jobs, accounting for a substantial proportion of economic growth (Carter et al. 2003).Furthermore, women-owned businesses also contribute to innovation, wealth creation and employment (Brush, de Bruin, and Welter 2009), Adopting the life course perspective, through linked lives, we investigate the gendered effects of household factors on routes into self-employment in the UK.In taking this approach, we also build on the work of entrepreneurship scholars who argue that we need to shift attention from the individual entrepreneur to the household context in order to understand how entrepreneurship is made possible (Carter et al. 2017) and that household and workrelated decisions and strategies are 'knitted together' (Alsos, Carter, and Ljunggren 2014) in partnership with significant others.Entrepreneurship scholars have examined the early antecedents of business start-up and how the factors associated with entrepreneurship differ across the life course (Jayawarna, Rouse, and Macpherson 2014).Our study is of the wider phenomenon of self-employment and we also focus on entrepreneurship separately, in both cases studying how the outcomes of individuals within their households depend on key life course events, rather than taking either a purely individual or household level of analysis.Studying this topic is important, because understanding the gendered nature of linkages deepens what we know both about self-employment, entrepreneurship and gender inequality in the labour market and how these relate to one another.Thus, we address two research questions which pursue the life course perspective on linked lives: (i) How do key transitions within the household context impact the likelihood of men and women moving into self-employment from employment?
(ii): How do the linked lives of individuals in couples impact self-employment transitions?
In the next section we outline the flexibility rationale for self-employment, then using theoretical ideas of linked lives we explain how one partner's self-employment transitions in a couple relate to the other partner's long hours.We examine previous research on the effect of changing relationship status through cohabitation, getting married or divorced, key turning points in the life course (Walther, Stauber, and Settersten 2022).We then examine these issues in relation to the subset of those individuals owning their own businesses to test if there are substantial divergences in relation to the key variables within the category of self-employed.We describe the data from Understanding Society, the variables we use and explain our analytical strategy which employs fixed-effects models.We then present the results and discuss our findings.

Background: gendered household transitions to self-employment
We know that women who face obstacles in reconciling work and care (because care tends to be normalized as their responsibility) often find that self-employment and entrepreneurship can provide a viable alternative occupation, although sometimes of lower quality (Budig, 2006a(Budig, , 2006b) ) (Georgellis and Wall 2005;Thébaud 2015).;In the following literature analysis, we showcase the main arguments in the scholarly works on gender, self-employment and entrepreneurship by looking at the promise of flexibility and autonomy for self-employment, and how specific episodes in life courses may interject to produce gender specific work transitions for linked partners in a household: arrival of children, recompensatory mechanisms of gendered relationality, marriage and co-habitation, and other factors.

Flexibility rationale
The transition to self-employment from employment is a significant development in the life course for both men and women, potentially presenting a plethora of opportunities for individuals to take charge of their professional and financial trajectory.Studies identify flexibility and autonomy as core motivations for transitions to self-employment (Barnett and Bradley 2007;Boden 1999;Smeaton 2003) for both men and women, as self-employment may provide the freedom 'when, where and how to work' (König and Cesinger 2015, p. 532).In popular accounts, self-employed entrepreneurs seem to have more control over their work hours, schedules, pace and location (Alsos, Carter, and Ljunggren 2014;Clark 2000;McGowan et al. 2012;Smeaton 2003;Thébaud 2015).For women, that motivation for freedom is often driven by linked lives and specifically the need to care for children (Brush, de Bruin, and Welter 2009;Thébaud 2015) or to accommodate work and life concerns (Carter et al. 2015).The work -family conflict literature presents flexible work arrangements as a potential resolution to conflict, arguing that job autonomy and schedule flexibility reduce conflict, particularly for women (Chung 2011, Russell, O'Connell, andMcGinnity 2009;Hill et al. 2001;Hofäcker, König, and Gregory 2013).However, care is needed in pursuing this argument as flexibility per se is not always synonymous with autonomy, particularly in employment (Henley and Lambert 2014).

The arrival of children
Against the institutional and social framing that caring work pertains to women, mothers may use self-employment to provide flexibility which allows them to maintain some kind of paid employment, a similar flexibility role to part-time employment (Georgellis and Wall 2005).This flexibility may help parents to meet childcare needs (Carr 1996;Kirkwood and Tootell 2008) in the face of intense childcare constraints, certainly in the UK where costs are high and the childcare service is of inconsistent quality (De Henau 2022).Related to childcare, working at home and home-based small business ownership are much more prevalent among women than among men in the UK (Felstead and Jewson 1997;Thompson, Jones-Evans, and Kwong 2009).This motivation for flexibility in location also underlies the rationale for women setting up their own businesses (as in Budig 2006a for the US and McGowan et al. 2012 for Northern Ireland).It may be that, as with men, compensating differentials are at work in that women are willing to accept lower pay or progression prospects in return for flexibility.Scholarly works also show that not having childcare responsibilities has also been associated with business start-up in the UK (Jayawarna, Rouse, and Macpherson 2014), potentially a different kind of start-up activity (Thébaud 2015).
Thus, the lives of individuals are interconnected in households and families particularly around providing care.Interconnections mean that the decisions of one household member, for example in work, can impact others in the family (Carr 2018).The linked lives perspective emphasizes the multifaceted nature of the decision-making process, emphasizing that outcomes are not determined only by individual factors, but involve other members of the household, as specifically elaborated in relation to small business (Brush, de Bruin, and Welter 2009;Carter et al. 2017).We investigate whether interconnections between family and work differ between men and women at key turning point through hypotheses H1-H3.We further explore the separate dimension of business ownership, which we expect to differ from self-employment overall, in H4.
Having children clearly makes a difference to joint working outcomes within couples in the UK in that in 56.5% of couples with a child under the age of three the father works full-time and the mother works part-time.In marked contrast, where the youngest child is in the age range 16-18, both parents work full-time in 58.3% couples (ONS 2019).Jayawarna, Marlow and Swail (2020) identified women with younger children as more likely to exit business ownership than women with older children, which they attributed to greater pressures in reconciling the demands of running a business and looking after children.It is entirely possible that having children prompts both entry to and exit from self-employment for women, consistent with women's multiple work transitions when children are very young, which partly also depend on having subsequent children (Stewart 2014).
It is likely that birth parity plays an important role which is largely neglected in most studies that average the states of having or not having children (Doren 2019).The arrival of a first child in a household is a flashpoint that triggers gender inequality, for example, enhancing the unequal division of paid and unpaid work as Musick, Bea and Gonalons-Pons (2020) found across the US, Germany and the UK, although Doren (2019) found that in the US women who only ever have one child are less likely to leave the workforce.
Based on these theoretical arguments about flexibility and having children, we posit the following hypotheses: H1: Becoming a parent makes the transition from employment to self-employment and business ownership in the UK more likely for women but not more likely for men (gender flexibility argument) H2: Women with one child are less likely to transition to self-employment than those with multiple children (birth parity argument).

Gendered relationality within opposite-sex couples
As we have set out, Moen (2003) and Moen and Sweet (2003) theorize that within couples, each partner's individual path is partly the result of the joint decisions couples make, which in turn are shaped by what is available within work organizations, social expectations of people and individual preferences.As Moen (2003) and Moen and Sweet (2003) indicated, if both partners in a couple with children work long hours, there is little time for children.Although not usually seen in this way, Gary Becker's (1991) specialization thesis is a conceptualization of linked lives through the male breadwinner-female homemaker model, a result of household efficiency maximization in the presence of discrimination.
Our article adds a new dimension to the study of such linkages, by examining the effects of an individual's working hours on their partner, within opposite sex couples.Paid working hours are a key source of gender division (Bryson et al. 2020).Time availability for paid work is determined by a couple's joint and individual decisions.Cha (2010a)'s study found that men's long hours of work are associated with women leaving work and couples moving to a male breadwinner -female homemaker household model.When women represent the main or equal earners, their male partners on average work fewer hours than other working men, potentially a sign of their weaker labour market situation (Kanji 2013).For men working fewer paid hours than a female partner is reflected in lower levels of wellbeing, illustrating the power of conformity to normative gender roles (Gash and Plagnol 2020).Further, research shows that working longer hours than a partner is associated with lower life satisfaction for women but not for men (Fleche, Lepinteur, and Powdthavee 2020).A similar type of within couple issue is whether a man's long hours of work lead to a female partner's transition to selfemployment as a way of achieving greater flexibility and autonomy to remain in paid work, rather than fully exiting paid employment.In order to further our understanding of how the couple's context differentially impacts the transitions of men and women to self-employment we propose H3.

H3:
Long working hours of a male partner prompt the transition to self-employment for mothers (gendered and linked lives argument).

Marriage and cohabitation
The life course perspective predicts that institutional contexts such as marriage versus cohabitation, are associated with differential outcomes in country-specific ways (Perelli-Harris and Lyons-Amos 2016).In the UK, cohabitation accounts for 16.3% of families with dependent children, while married and civil partner couple families account for a larger share of 61.4%.Differences between marriage and cohabitation relate to legal rights to the other's assets and relationship stability, in that cohabiting couples with children have been found less likely to stay together than married couples (Kanji and Schober 2014;Kiernan and Mensah 2010).Taken together these factors are likely to affect how couples perceive the strength of risk sharing in marriage versus cohabitation.
In accord with this prediction, Özcan (2011) found in the US context that marriage -and, to a lesser extent, cohabitation -for men are associated with a transition to self-employment overall and there is an even stronger association between marriage and business ownership.One study in the UK reported that married men are more likely to be self-employed (Atherton, Wu, and Wu 2018).Özcan's (2011) findings are different for women in that only marriage, and not cohabitation, is associated with a transition into unincorporated self-employment although not with the transition to small business ownership.
The disparities between the effects of marriage and cohabitation on the transition to business ownership and the broader category of self-employment might be due to divergences in wealth between married and cohabiting people (Perelli-Harris and Lyons-Amos 2016) and between business owners and the self-employed (Mwaura and Carter 2015).Furthermore, self-employed individuals overall may differ from the subset of business owners (Dawson and Henley 2012).We examine business ownership separately from self-employment overall because household contexts are particularly important for small business growth (Carter et al. 2015;Alsos, Carter, and Ljunggren 2014;Mwaura and Carter 2015b;Carter et al. 2017) and business exit decisions (Jayawarna, Marlow, and Swail 2020).
An alternative route of influence, based on support, occurs for women in more egalitarian couples (Hook 2010;Thébaud and Pedulla 2016), which in the UK applies to cohabiting couples (Kanji and Schober 2014).In more equal relationships partnered mothers could be more likely to engage in running their own business if the partner shares parental responsibilities related to childcare and housework, as has been found in the more egalitarian context of Sweden (Naldi et al. 2021).Women are less likely to enter self-employment in more egalitarian societies, but when they do, they start ventures that are more in line with high-growth-path businesses (Thébaud 2015).In this scenario, we might not expect the arrival of children to push mothers into self-employment for flexibility reasons, but an equal partner might support women in business ownership rather than push them into it.This line of argument is similar to Verbakel and de Graaf's (2009) thesis that the social capital between members of a couple enhances the probability that they will both engage in paid work.
These arguments lead us to the following hypothesis: H4: There is a stronger association between marriage and business ownership than between marriage and self-employment and this relationship is stronger for men than women (risk pooling argument).

Other factors influencing self-employment
Our interest is in linked lives, specifically how demographic life course transitions such as entering marriage and cohabitation, having children of different birth parities, and couples' joint work arrangements precede or coincide with transitions to self-employment.As such, our analysis focuses on the effects of time-varying variables on within person change, as we discuss in the next section.Previous studies of transitions to self-employment have found that time-varying economic factors such as the state of the economy and access to credit (Kepler and Shane 2007;Nanda 2008) influence both men's and women's transitions to self-employment (Henley 2017;Saridakis, Marlow, and Storey 2014).Amongst the many studies of time-invariant factors (not the focus of this study), findings are that already wealthy individuals (Nanda 2008), with a family background in business (Colombier and Masclet 2008), from more advantaged social class positions (Jayawarna, Rouse, and Macpherson 2014), and those from minority ethnic backgrounds (Brynin, Karim, and Zwysen 2019) are more likely to become business owners.Becoming self-employed is associated with enhancements in job satisfaction (Benz and Frey, 2008) with the causal pathway running from self-employment to job satisfaction rather than job satisfaction causing self-employment.Furthemore, with regards to self-employment, studies found differences in performance outcomes in the US being linked to a lack of experience in family businesses and by race (Fairlie and Robb 2007).Self-employed workers are on average of a higher age than the employed (Simoes, Crespo, and Moreiro, 2106).In addition, better health status (Rietveld, Van Kippersluis, and Thurik 2015), degree of risk aversion and a partner's involvement in either employment (as a risk mitigation strategy) or in self-employment have all been found to be associated with self-employment (Simoes, Crespo, and Moreira 2016).

Data
To test the hypotheses, we use the UK's Household Longitudinal Study Understanding Society (USoc), a large-scale, nationally representative household panel survey covering all members of the sampled households, which allows us to match partners to each other.This possibility permits the study of household factors and how interrelations within the household affect being selfemployed or making the transition to self-employment over a sustained time period.We used 10 waves of the survey, which led up until the COVID-19 pandemic started in the UK.Sample waves were collected over two-year periods, with a new wave starting annually.
The analysis uses data from waves 1 to 10: Wave  2018-2020 (n = 34,318).The Wave 1 sample has characteristics that are very similar to the characteristics of the UK government's Labour Force Survey (Lynn and Borkowska 2018).The survey was designed to be nationally representative through a multistage sample design, mostly with stratification and clustering (Buck and McFall 2011).
As in most large-scale longitudinal surveys, attrition between survey waves took place.In Wave 1, the survey covered 40,000 households in Britain.The response rate among households sampled in Wave 1 was 57.6%.Of the individual respondents in Wave 1, 64.5% were re-interviewed at Wave 2 (Lynn et al. 2012).The panel is unbalanced in that we do not have complete histories for each individual.In the models with lagged period variables, we included all observations for individuals with three consecutive years of data.For contemporaneous models, we included all observations on individuals with two consecutive years of information.Demographically, we focused on the main working ages between 21 and 55, following Özcan (2011).Georgellis and Wall (2005) set the later age of 59 as the cut off age in their study.Below the age of 21 many are still in education.After the age of 55, people are known to change their work behaviour, which for some includes entering selfemployment as bridge employment between employment and retirement (van Solinge, 2014; Zissimopoulos and Karoly, 2009).The factors associated with this later stage transition are rather different from stages in the life course when men and women have children, both stages are important but require separate treatment (Cheraghi, Wickstrøm, and Klyver 2019).The distribution of self-employment is skewed towards older ages as Table 1 shows, but a study in the USA found that about a third of the self-employed made the transition after the age of 50 and the remaining two thirds of the transitions were at younger ages (Zissimopoulos and Karoly, 2009), Most people make the transition to being a parent of a child of any parity by the age of 55, although it is technically possible to have children after this age.Models are estimated separately for men and women, avoiding the potential bias from assuming that all other variables have the same effect on men and women (Brush, de Bruin, and Welter 2009;Sprague 2005).

Dependent variables
We used three dependent variables in our estimations.The first dependent variable is a dichotomous variable that measures whether an individual is in paid employment or self-employment.In Understanding Society, the self-employed include those who are running a business or professional practice, partner in a business or professional practice, working for themselves, serving as a subcontractor, doing freelance work or being self-employed in some other way.The second dependent variable is coded as 1 if the individual is a business owner (that is an answer in the affirmative to running or being a partner in a business or professional practice) and 0 if he or she is employed.In the final part of the analysis, the third dependent variable is coded 1 if individually employed and 1 if self-employed (as above) but is restricted to the sub-sample of men and women within opposite-sex couples.The panel fixed-effects models assess changes in these statuses between years.

Independent variables
Family status is defined by a categorical variable defining whether the person is single, married, divorced or separated, a widow or widower, or cohabiting as a couple.
Number of children is the number of children under age 16 within the household; the reference category is zero.
Education is measured by the highest level of education attained in ascending order: no qualifications, other qualifications, GCSE-level qualifications or the vocational equivalent, A-level qualifications or their equivalent, other higher education (not degree level) and degree-level qualifications.
Age and age squared together act as a proxy for relevant accumulated experience, which reflects the proposition that the impact of experience is likely to be non-linear.
Long hours of work are constructed from the question asked directly of and answered by the partner on his or her number of hours normally worked per week: The base category is less than 48 hours per week.The variable is coded 1 if the partner works 48 hours or more per week (robustness checks have also been performed by slightly amending the hours cut off).

Control variables
Residence in the different regions is controlled using a 12-category variable that divides the UK into the North East, North West, Yorkshire and Humber, East Midlands, West Midlands, East of England, London, South East, South West, Wales, Scotland and Northern Ireland.The variable for region captures differential rates of growth across regions over time.We also constructed and employed a separate variable for regional unemployment rate using ONS reported rates.In addition, we controlled for changes in health status prior to commencing self-employment using the variable recording satisfaction with health (sclfsat1).

Analytical strategy
We examined within-person changes over time using fixed-effects panel logistic regressions.As Allison (2009) explains, fixed-effects model how deviations from an individual's group mean in the dependent variable relate to changes in an individual's group mean in the explanatory variable.Fixed effects are the method of choice in many studies evaluating the causal effects of becoming parents on people's lives (Perales 2019), an aim relevant to the analysis of the effects of marriage, partnership and having children to investigate H1 and H2.Fixed-effects models reduce omitted variable bias in respect of time-invariant variables (for example, personality or ability factors) that have not been identified but may affect the likelihood of an individual transitioning to selfemployment.Time-variant unobserved variables could still act as factors that induce bias.However, most variables previously found to be associated with the transition to self-employment are time invariant, such as whether a person's father or mother was self-employed.Although fixedeffects models are inefficient because they ignore between-group variation, they are consistent.
In the random effects model, unobserved time-invariant characteristics are assumed to be zero, otherwise the estimator is inconsistent.As a robustness check, we employed the Hausman test which leads us to reject the null hypothesis that unobserved individual-level effects are uncorrelated with the other covariates, for the male equation (chi2(24) = 60.72 p > chi2 = 0.000, and for the female equation, chi2(24) = 60.83,p > chi2 > 000).This result implies that we should use the fixed-effects estimator instead of the random-effects estimator.
We estimated the models in Stata 15 using the xtlogit and clogit procedures (which are essentially the same) with robust standard errors.The effects are fixed for persons and for years by including dummy variables for years.Regarding H2, we separately examined whether the transition to parenthood is associated with a higher likelihood of entering self-employment one year after the birth of a child using variables lagged by one year (t 2 -t 1 ) and variables that are not lagged (t 1 -t 0 ), a robustness check for the change in the explanatory variable coincides with the change in the dependent variable (association rather than causation), which provides greater insight into H2.As stated, we examined births by parity.
We are mindful of the risk of including variables that are the consequences rather than the causes of self-employment (see Georgellis and Wall 2005), which is why we included lagged variables and did not include variables such as job satisfaction.Non-random selection into any kind of paid work is possible, but during the period we examined, 73% of women in the sample were employed or selfemployed, compared with 84.81% of men.
Relating to H3, our interest is the household context concerning whether individuals in a couple are more likely to make a transition to self-employment if their partner works long hours (over 48 hours per week).Confining our analysis to opposite-sex couples (as the sample size for same-sex couples is too small) in which the partner works, we explore how one partner's long work hours affect the other partner's transition to self-employment separately for men and women (Model 3).

Results
Table 1 presents summary statistics applying weights provided by Understanding Society and using Stata's svy commands to adjust the sample to represent the population.We used crosssectional weights for each round of data and then averaged the summary statistics across the 10 years of weighted data.Table 1 shows substantial differences between self-employed men and women in the 21-55 age range (overall women's participation in self-employment is at a lower level).A clear difference is in the highest qualification: On average self-employed women have a much higher level of educational attainment than self-employed men: 40.64% have degree-level qualifications compared to 29.26% of self-employed men (t-test of the difference is significant, p < 0.01), and it is also higher than the 37.25% of employed women (t-test of the difference is significant, p < 0.001).In contrast, self-employed men tend to be disproportionately represented in the lowest and lower levels of educational attainment compared to employed men.A fairly high proportion, men and women, both self-employed and employed (58% of employed men and 54.37% of employed women), do not have children under the age of 16 living in their households, although they may have children older than 16.Male business owners are much more likely to have children than either self-employed or employed men.
Male and female self-employed and business owners are less likely to be single than the employed; there are similar levels of cohabitation across self-employed and employed, although a lower proportion of self-employed women cohabit, having a substantially higher probability of being married.Only 11.30% of male business owners were single, compared to 19.57% of selfemployed men, the difference is statistically significant, p < 0.01.Table 1 points to regional variations that require further study using national data: most notable for our study is that a disproportionate (in comparison to women in employment) share of women in self-employment and business ownership are located in London and, even more so, the South-East.
In Table 2, Model 1 shows that for men, the transitions to marriage and cohabitation from being single were not associated with an increase in the odds of moving from employment to selfemployment, either in the lagged (Model 2) or current period models (Model 1).Our results for women are more in line with previous studies in that there is a weakly significant impact of marriage on the likelihood of making a move to self-employment (1.46 times the odds), but not of cohabitation (models 1).For women, the marital status variable lagged by one period was also significant at p < 0.05 (OR 1.68).The significance of the lagged variable for married women makes us reasonably confident in making statements about the causality in transitions to self-employment from employment.Regarding business ownership (Model 3), the results suggest a stronger association between marriage and business ownership than marriage and self-employment, in support of H4.Moving from being single to married has a large impact on both men's transition to business ownership (2.78 times the odds) and women's transition to business ownership (3.48 times the odds), significant at p < 0.01 in both cases.Similarly, cohabitation relative to being single raises the odds of both men (1.73 times the odds) and women (1.97 times the odds) making the transition to business ownership, although it is a lesser magnitude of association than for marriage.Interestingly, divorce is associated with men becoming business owners, but this is not the case for women.
We set out H1 in relation to our expectation on the effects of children.The findings support the hypothesis that becoming a parent does not affect the likelihood that men make a transition to self-employment or business ownership from employment: in both lagged and current period models, one, two or three or more children are not significant.In relation to H1 for women, the results from Model 2 reveal that the transition to having a first and second child significantly raises the odds of them moving to self-employment from employment, which is the current period effect.Women's transition to having a first or second child significantly raises the odds of women moving from employment to selfemployment and separately to business ownership.In the lagged model for self-employment, we see that only moving from having a first to a second child significantly raises the odds of women making the transition to self-employment (by 1.59).These results support H2 which anticipates differential effects by birth parity for women.As far as women's business ownership is concerned, first and second child births raise the odds of women moving away from employment.We interpret the lack of the effect in relation to having a third child for the transition to self-employment and business ownership due to the associations with earlier birth parity.A higher proportion of women who are self-employed (7.92%) and business owners (9.00%) have three children than employed women (4.8%) but it seems that they make the transition to self-employment before the birth of the third child.Decisions around employment status and fertility behaviour are related and, from our results, based on life course and linked lives perspective also sequenced.
The summary statistics in Table 1 show that self-employed men are a little more likely than employed men to have co-resident children in the household, but the results presented in Table 2 show that having a first, second or third child is not associated with the transition to selfemployment or business ownership for men.
In H3, we were interested in evaluating whether one partner's long hours at work might prompt an individual to move to self-employment to accommodate the partner's longer hours.We chose 48 hours as the cut-off for long hours as this is the maximum permitted number of hours of work under the Working Time Directive 1998 with the exception of those who voluntarily opt-out or work in certain specified occupations.Full-time work in the UK is usually classified as 35 hours or more per week, while previous research has found that men in the UK work among the longest hours in Europe (Author citation).The results, as presented in Table 3, add to our knowledge of the transition to self-employment by showing that when a male or female partner works over 48 hours per week, an opposite sex partner is more likely to make the transition to self-employment.Somewhat against the expectations we set out in H3 this situation applies to both men and women but women are much more likely to be in this position than men (6 times more likely in the sense that amongst the couples in the sample 9.3% of women have a partner working long hours but only 1.5% of men are in this position).This difference in the proportions of men and women in the position of having a partner who works long hours makes the effects asymmetrical.Looking separately at Coefficients are odds ratios.***p value < 0.01,**p value < 0.05, *p value < 0.10.Controls for education, wave, region and regional unemployment included.In the male equations (models 1 and 2), the wave dummies are not significant, but all wave dummies after Wave 2 in the female equations for models 1 and 2 are positive and significant at the 99% or 95% level.
business owners, as Table 3 shows we did not find that a partner's long hours prompted the transition to business ownership for either men or women (results not reported).This finding tends to confirm that business ownership differs from self-employment overall.

Discussion
The application of the concept of 'linked lives' presented by Moen (2003) and Moen and Sweet (2003) led us to examine the transition to self-employment as a turning point in employment trajectories related to life course events.The transition is brought about through negotiation, decision making or the operation of societal norms (the macro context in Brush et al. (2009)'s model) which take effect within and outside the household, carrying implications for the entire family.Our results demonstrate how this transition is associated with the demographic turning points of entering marriage, cohabitation and having children, and the joint gendered strategies linked individuals make, which represent the key elements of the life course approach (Elder 1994;Settersten 2015a).Using nationally representative data for the UK, we were able to estimate the causal effect of entering relationships such as marriage and cohabitation, and having children by different birth parities, on the odds of entering self-employment.While previous studies identified that situations within households, such as having children and being in partnership, were associated with selfemployment (Budig 2006a(Budig , 2006b;;Carr 1996;Özcan 2011;Saridakis, Marlow, and Storey 2014), few studies have examined this relationship using fixed-effects, the method of choice in studies of the effects of parenthood on people's lives (Perales 2019).We have also drawn limits around our claims of causality using the same period and lagged variables.
Couple-level arrangements are not solely about maximizing resources but are also driven by men and women enacting gender roles through performance (West and Zimmerman 1987).We interpret the results as demonstrating this performance.In particular the arrival of children has asymmetric effects, associated with women's moves to self-employment, a result that is not seen for men.A robust finding from the lagged and current period models is that a second child has a causal effect on women's transition to self-employment, reinforcing the need to look at children in the context of fertility over time, something rarely examined or discussed in scholarly works.Existing research across contexts, including the UK, tends to examine children as a category rather than by birth parity (Atherton, Wu, and Wu 2018;Clark 2000;Naldi et al. 2021;Smeaton 2003;Thébaud 2015).An exception is Stewart's, 2014 research in the UK that shows that birth parity matters for women's flows out of employment.Our results show the effect of the arrival of a second child is stronger than a first child.This is an important finding in view of Table 1 which shows that a higher proportion of women who are self-employed and business owners have two or three children than is the case for women in employment.For example, 4.80% of women in employment have three children, compared to 7.32% of self-employed and 9.00% of business owners (the differences are statistically significant at p < 0.01).These findings enable us to offer a more nuanced and sequenced view of how children affect occupational transitions in households in the UK.

Working hours within couples
We set out previous research on the relationality of employment paths which requires the study of interactions at the couple or household level.An important finding about our research question and hypothesis (H3) relating to 'linked lives' (Moen and Sweet 2003) is that, for opposite-sex couples, a transition to long work hours of a partner significantly increases the odds of the other partner making a transition from paid employment to self-employment.This result is similar to Cha's (2010b) finding that women withdraw from work due to a male partner's long work hours, but notably different to her finding that a male partner does not withdraw from work because of a female partner's long hours.Indeed, a number of studies of couple interaction, find asymmetric gendered effects in terms of wellbeing, particularly when gendered norms are contravened (Fleche, Lepinteur, and Powdthavee 2020; Gash and Plagnol 2020).It is therefore notable that the analysis in this article shows that a woman's long hours of work also prompt a male partner's transition to selfemployment.However, we must interpret this seemingly symmetrical finding about the effects of a partner's work hours in relation to the scale of those affected: still women are far more likely to be in a situation where their partner works long hours (9.3% of women were in this position compared to only 1.5% of men) which makes this process highly gendered because women are much more likely to move to self-employment because of a partner's long hours.Self-employment may, in this situation of a partner's long hours, act as a buffer zone between not working and being in paid employment, fulfilling a similar role to that performed by part-time work in the UK (Georgellis and Wall 2005).

Business ownership is distinct from self-employment overall
The findings illustrate the importance of marriage to the transition to business ownership for both men and women in the UK.This association accords with the risk sharing argument, that having a partner can help to mitigate income risks or that a partner may contribute by providing emotional support, consistent with Naldi et al. (2021).As the transition to marriage is associated with an increase in the odds of moving to business ownership, it seems likely that decisions about entering business are facilitated by relationship commitment.Numerous studies document how marriage is associated with higher employment status and wages for men.In line with the overall findings for the wider category of self-employment, children are associated with women's transition to business ownership while there is no sign that men make the move to business ownership in association with having children.This differential association between arrival of children and the gendered transition to self-employment could add to the evidence of low gender egalitarianism in the UK.Moreover, it could imply that women's pathway to entrepreneurship is constrained (Cheraghi, Wickstrøm, and Klyver 2019), if women are making the transition to bring flexibility to accommodate work and care.
Turning to the limitations of our study, we chose fixed-effects models to gauge causal effects, but this did not permit us to examine variables that are not time-varying but are associated with selfemployment in other studies.We were able to demarcate business ownership within selfemployment, but we could not distinguish the distinct, high-growth and high return-oriented activity of entrepreneurship.More targeted surveys are required for this kind of quantitative research.Simultaneously, qualitative research in this area would permit more nuanced investigations of relationality and participants' own interpretations of their actions and situation.Our panel data study adds to our knowledge of the factors associated with business transitions, contributing the business entry perspective to Jayawarna, Marlow and Swail's 2020 study of business exits.

Conclusion
This study of transitions to self-employment provides a view into a gendered system, and making explicit the part that self-employment plays in this system.These work transitions take place at specific points in the life course, shaping interconnections in the lives of members of a couple and between parents and children.Using the linked lives lens, we examined how key turning points such as marriage, cohabitation, having children and a partner's move to long working hours influence how these linkages are performed.The births of children represent a key turning point with many ramifications for ways of organizing, and a correlate of women's inferior labour market position in the UK (Costa Dias, Joyce, and Parodi 2020) and elsewhere.Our study revealed that children and birth parity are consequential for transitions into selfemployment, underlining the importance of perceiving children separately by birth parity, as Doren (2019) has argued.
In the broader labour market terms, women's moves to self-employment are often excluded from studies of inequalities.While past studies have recognized and accounted for the selection effects in women's employment, they have excluded the self-employed in evaluating such selection effects.We would argue that, to understand women's overall status in the labour market, we also must understand what makes them move into and out of self-employment, the types of activities they engage in as self-employed workers (Budig 2006a) and whether the move to self-employment represents occupational downgrading to achieve flexibility similar to the move from full-time to parttime work (Connolly and Gregory 2008;Georgellis and Wall 2005).
We know that dropping out from employment is highly consequential for women's longer-term work-related outcomes, but we have little information about how periods of self-employment or the continuation of self-employment affects women's longer-term outcomes, for example in, relation to older age income provision.Indeed, taking a life-course perspective and linked lives lens on moves between employment and self-employment is critical to understanding the enduring effects of these transitions into retirement and older age.Further study will be needed to investigate this important issue, which informs our overall understanding of women's position in the labour market.Moreover, research is needed to understand the consequences of transitions to different types of selfemployment.Further research could uncover the variety of household influences on linked lives beyond the ones considered in this article and at different and turning points and stages of the life course, which generally offer a more holistic and intertemporal view of the household as a site of gendered work and life-related experiences.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Table 2 .
Fixed-effects models: the effect of marital status and children (models 1 and 2) on the likelihood of moving from employment to self-employment, and business ownership (model 3).

Table 3 .
Fixed-effects models: the effect of a partner's long hours (model 3) on the likelihood of moving from employment to selfemployment.