Executive summary

In Australia, as in several other OECD countries, the proportion of lone mothers participating in paid work is lower than that of couple mothers (Gray, Qu, Renda, & de Vaus, 2006; McHugh & Millar, 1996; Millar & Evans, 2003). The lack of employment in lone parent households contributes to concern about the wellbeing of adults and children living in jobless households, and is therefore of considerable policy interest.

This paper explores differences in lone and couple mothers' employment by examining how transitions into and out of employment for these two groups vary. A focus on transitions enables us to examine whether the lower employment rate of lone mothers is due to their being less likely to enter employment or more likely to exit employment once they are employed (or a combination of the two).

Month-to-month employment transitions of lone and couple mothers are examined using the employment calendar data from Waves 2 to 8 of the Household Income and Labour Dynamics in Australia (HILDA) survey, covering the period from 2001 to 2008. HILDA is a nationally representative, annual panel survey, which commenced with Wave 1 in 2001. The analyses include mothers of children aged less than 15 years, which was approximately 2,000 persons at each wave.

The calendar component of HILDA is obtained at each wave of the survey, with details of work activities being collected for the preceding year. These data allowed derivation of mothers' employment status in each month. Details of mothers' relationship status could also be derived at this monthly level, and matched to a range of other details, such as ages of children and prior work history.

Initial analyses of mothers' employment transitions revealed that the vast majority of employed mothers remained employed from one month to the next, and similarly the vast majority of not-employed mothers remained out of employment. Averaged across all the data, 1.3% of employed couple mothers and 2.0% of employed lone mothers left employment from one month to the next. The exit rate from employment was, therefore, higher for lone than couple mothers. While these rates are small, over longer time periods (e.g., over one year) they translate into significant movements out of work, especially for lone mothers. The monthly transition rates into employment were also small in size, and here there was no evidence of differences between lone and couple mothers, with both having average rates of 3.6% of not-employed mothers taking up employment in the subsequent month.

After comparing the employment transitions of lone and couple mothers at the aggregate level, comparisons were also made across the following characteristics: recent work history (capturing the percentage of the previous year in employment), educational attainment, age of youngest child, number of children, health status, country of birth and age. These variables were of particular interest given that lone mothers were expected to have lower levels of human capital than couple mothers, and to face greater barriers in entering or remaining in the labour market. A comparison of these characteristics in these data confirmed this was the case.

Using descriptive techniques, these analyses showed that some factors were especially important in explaining variation in employment transitions. For both transitions into and out of employment, recent employment history was one such factor. Those who spent a longer part of the previous year in employment were more likely to remain employed, if employed, and more likely to enter employment, if not employed. Level of educational attainment was also an important factor, with higher education associated with lower rates of movement out of employment and higher rates of movement into employment. Significant differences were also apparent for age of youngest child, number of children, self-reported health status, and country of birth of mothers.

Multivariate analyses were used to examine the variation in employment transition rates according to these different characteristics, and to examine the extent to which differences in employment rates of lone and couple mothers were apparent after taking account of this range of characteristics. Random effects models were used, given the repeated (monthly) observations for each person. The multivariate analyses showed that the difference in rates of leaving employment, for lone and couple mothers, remained apparent after taking account of their different characteristics. That is, the higher rate of leaving employment for lone mothers was not entirely due to differences in observed characteristics of lone and couple mothers. The multivariate analyses of transitions into employment confirmed the initial descriptive results: that there were no differences in movements into employment for lone and couple mothers.

Variables examined in the descriptive analyses were also important in the multivariate analyses. The significance of variables such as recent work history and educational attainment in explaining variation in transitions into and out of employment is relevant when considering lone and couple mothers, given that lone mothers, on average, had less recent work history than couple mothers, and lower levels of educational attainment.

In addition to these analyses of HILDA calendar data, the report incorporates information collected about the labour market characteristics of lone and couple mothers at each survey. For employed mothers, this includes more detailed information about job characteristics; for not-employed mothers, this includes more detailed information about employment preferences, barriers and work history. These data show that employed lone mothers were more likely than couple mothers to be in casual employment, and to have entered or exited employment during the year. For not-employed mothers, lone mothers are more likely than couple mothers to be actively seeking work and available to start work; however, they also appear to have more barriers that may impede their take-up of employment. For example, in most years, lone mothers were more likely than couple mothers to say their ill health, injury or disability was a reason for their not looking for work, if they were not in the labour force.

From a policy perspective, these analyses reinforce the importance of employment retention, which is more of an issue for lone than couple mothers, given their higher exit rates from employment. Potential policy approaches for those in employment may include the provision of supports by case managers, training or coaching. For mothers who are not employed but seeking employment, tailored assistance with job search and access to further education or training is likely to be beneficial.

Any policy approach needs to be considered in relation to the structure of the labour market and the availability of suitable jobs, as well as the availability of non-parental child care. Further, the particular characteristics of mothers need to be taken into account, and solutions should be developed for mothers with lower levels of education, limited work experience or health issues. For some mothers, especially those with very young children, very different policy needs are likely to exist, given the importance of the caring role for mothers.

To summarise, we found that, compared to couple mothers, lone mothers had a somewhat higher rate of exiting employment in any one month, but no difference emerged between lone and couple mothers in their rate of entering employment. The different rates of exiting employment remained significant when certain characteristics of lone and couple mothers were taken into account. Policy measures aimed at increasing the job retention of lone mothers may therefore help to increase the employment rate of lone mothers. However, the vast majority of not-employed lone mothers are remaining out of work rather than transitioning between non-employment and employment. Some of these mothers are likely to be focusing on providing full-time care to their children at this life cycle stage. Others, particularly lone mothers, want to work, but face substantial barriers to doing so. Providing additional tailored supports to these mothers may be another way of reducing the employment gap between lone and couple mothers.

Introduction

While the proportion of lone mothers participating in paid work in Australia has increased over recent years, this proportion remains below that of couple mothers (Australian Bureau of Statistics [ABS], 2009; Gray Qu, Renda, & de Vaus, 2006; McHugh & Millar, 1996). The same is true in other Organisation for Economic Co-operation and Development (OECD) countries, such as the United Kingdom (UK), the United States (US) and New Zealand (Millar & Evans, 2003). Concerns about the wellbeing of adults and children living in jobless households contribute to continuing interest in explaining the relatively low employment rate of lone mothers.

Many explanations have been given for lone mothers' lower rate of employment,¹ including differences in lone and couple mothers' characteristics, wages and abilities to combine work with caring for children, including differential access to informal child care networks (Eardley, 2001; Gray et al., 2006; Hynes & Clarkberg, 2005; McHugh & Millar, 1996; Walters, 2002a). The role of government support is also important, especially as lone mothers are more likely to be in receipt of income support payments and are thus more likely to face financial disincentives to work due to the interaction of the income support system with wages (Adema & Whiteford, 2007; Millar & Evans, 2003).

This paper provides new insights into lone and couple mothers' employment by examining how transitions into and out of employment for these two groups vary. A focus on transitions enables us to examine whether the lower employment rate of lone mothers is due to their being less likely to enter employment or more likely to exit employment once employed (or a combination of the two).

Month-to-month employment transitions of lone and couple mothers are examined using the employment calendar data from Waves 1 to 8 of the Household Income and Labour Dynamics in Australia (HILDA) survey, covering the period from 2000 to 2008. The HILDA calendar data provide a unique opportunity to analyse mothers' employment transitions over an extended period of time (seven years to date), with a sample large enough to be able to differentiate mothers according to their relationship status.

Throughout the paper, comparisons between lone and couple mothers are made. However, the main focus is on lone mothers, as for these mothers, lack of employment means family joblessness. Joblessness in families is linked to long-term welfare dependence, poor health for children and parents, lower levels of life satisfaction, housing and financial difficulties and poorer future educational and employment outcomes for children (Australian Institute of Health and Welfare [AIHW], 2008; Dawkins, Gregg, & Scutella, 2002; Heady & Verick, 2006). Also, extended periods of absence from the labour market while children are young can lead to later difficulties in workforce re-entry (Shaver, King, McHugh, & Payne, 1994).

Employed lone mothers are often in jobs that are short-term, unstable and low paid, which can lead to "churning" into and out of employment and reduced prospects for improved employment conditions in the longer term (Cancian, Haveman, Kaplan, & Wolfe, 1999; Chalmers, 1999; McHugh & Millar, 1996). In addition, some lone mothers receive little financial benefit from working, once child care costs, taxation and loss of income support and associated benefits and concessions are taken into account (Harding, Vu, Percival, & Beer, 2005; Toohey & Beer, 2003). Also, juggling paid work with child care may be more of an issue for lone than couple mothers, due to the absence of a partner to share child care responsibilities (Fok, Jeon & Wilkins, 2009).

Whiteford (2001) showed the importance of analysing the dynamics of lone mothers' employment, but no (published) studies of large-scale Australian datasets have focused on this topic. Several papers - including Barrett (2002), Chalmers (1999) and Gregory, Klug, and Thapa (2008) - have analysed differences by relationship status in the related topic of dynamics of income support receipt. Whiteford found that lone mothers were more likely to experience transitions into or out of a job over the period of a year than partnered men and women. He reported that a significantly higher proportion of lone mothers had a job during the year than were employed at any one point in time, indicating some mothers had periods during the year where they were unemployed or outside the labour force. Previous analyses of Australian data have shown that employment transitions vary with particular characteristics (such as age and prior employment experience; e.g., Knights, Harris, & Loundes, 2000; Stromback, Dockery, & Ying, 1998), but as they have not focused on mothers, they do not shed light on the question of how lone and couple mothers vary. One relevant study looking at employment transitions in the UK (Evans, Harkness, & Arigoni Oritz, 2004) found that in 2002-03 lone parents were moving into employment at similar rates to couple parents. However, they were almost twice as likely to leave their job as couple parents, despite the fact that the rate of job exit for lone parents had fallen between 1993 and 2003.

The structure of this paper is as follows. Section 1 provides contextual information on lone and couple mothers' employment in Australia and overseas, and includes a review of literature relating to mothers' employment transitions. Section 2 then presents an overview of the HILDA data and the methods used in this paper. Results are presented in Section 3, starting with an overview of employment transitions, and then more detailed analyses of leaving and entering employment are presented separately. Section 4 provides a discussion of the findings and policy implications, and concluding thoughts are given in Section 5.

1. Australia in context

1.1 Mothers' employment rates

The employment rate of lone mothers is lower than that of couple mothers in several OECD countries, including the UK, Ireland, Belgium and the Netherlands, as shown in Figure 1.² In other countries, such as Germany, France and the US, lone and couple mothers have similar employment rates, while in others, including the southern European countries of Greece, Italy and Spain, lone mothers have significantly higher employment rates than couple mothers.

Figure 1: Employment rates in selected OECD countries of mothers aged 15-64 years, by relationship status, 2007

Source: OECD Family database for all countries except Australia; Australian data sourced from ABS Labour Force Status by Sex, Age, Relationship (Supertable FM2), data for June 2007, for mothers of children aged under 15

In addition to diversity in societal expectations regarding maternal employment, these cross-country differences may in part be attributed to the varying levels of government support and associated work requirements for lone mothers, and different levels of in-work support, such as child care and case management (Adema & Whiteford, 2007; Eardley, 2001). The country-specific characteristics of labour markets, in particular the availability of part-time work, may also be important.

In Australia, the employment rate of lone mothers has, in recent decades, been lower than that of couple mothers, although the gap has varied across this time (McHugh & Millar, 1996). This is apparent in Figure 2, in which lone and couple mothers' employment rates are shown for June 1984 to June 2008. These data show similar rates of growth in employment until recent years, when the rate of growth for couple mothers has slowed, but for lone mothers it has increased, thus narrowing the gap between lone and couple mothers' employment rates.

Figure 2: Employment rates of mothers with children aged under 15 years, by relationship status, June 1984 to June 2008

Source: ABS Labour Force Status by Sex, Age, Relationship (Supertable FM2); ABS Labour Force Status and Other Characteristics of Families, 1984 to 2003

For the period under examination in this report, 2000 through to 2008, the employment rates of lone and couple mothers are shown by age of youngest child in Figure 3. These data show that lone mothers' gains in employment participation have been greatest for those with older children. In 2008, for example, for mothers with children aged 10-14 years, there was little difference in lone and couple mothers' employment rates.

Figure 3: Employment rates of mothers with children aged under 15 years, by relationship status and age of youngest child, June 2000 to June 2008 (ABS)

Source: ABS Labour Force Status by Sex, Age, Relationship (Supertable FM2)

The employment rates conceal another important dimension of labour force participation-the number of hours worked. While this is not covered in this paper, there are significant differences between lone and couple mothers in hours worked, and how the balance of full-time and part-time employment has changed. During the 1980s and 1990s, lone mothers' part-time employment increased significantly, while their full-time employment did not. As a result, in 2005, a higher proportion of employed lone mothers were working part-time than were working full-time. In 1997, the reverse was true (Gray, Qu, Renda, & de Vaus, 2006).

1.2 Government policy approaches

In Australia, historically, lone parents with dependent children have received income support not tied to labour market obligations, with a policy goal of alleviating poverty in those families, and recognising the parenting role of these lone parents (Cass, 2006; McHugh & Millar, 1996). By OECD standards, this assistance has been relatively generous (Whiteford, 2001).

While continuing to provide financial support to parents of young children, increasingly, government focus has shifted towards an emphasis on the importance of paid work (McHugh & Millar, 1996). By adopting policies that encourage lone mothers into employment, governments aim to improve lone mothers' future employment prospects, reduce the incidence of poverty, lower welfare costs and reduce the likelihood of intergenerational welfare dependency.

The history of assistance to lone parents goes back many decades (Cass, 2006; McHugh & Millar, 1996), but here we mainly focus on the period covered by our analyses of lone mothers' employment transitions, from 2000 to 2008. Prior to September 2003, lone mothers who met income and assets tests could receive a pension payment, Parenting Payment Single (PPS), until their youngest child turned 16 without having to fulfil any activity requirements (Centrelink, 2003).

The shift towards an emphasis on the importance of lone parental involvement in paid work began with the introduction of the Jobs, Education and Training (JET) program in 1989. This voluntary program for lone parents aimed to aid movement into employment, particularly for those who had been in receipt of benefit for long periods, and those nearing the end of their eligibility.

In September 2003, as part the Australians Working Together welfare reforms, many of the supports provided by the JET program were maintained and extended to all PPS recipients and, in addition, compulsory activity requirements were introduced. PPS recipients whose youngest child was 13-15 years old were required to undertake one or more activities (such as job search, education, training or community work) for up to 150 hours in each six-month period to develop and enhance their work-related skills to prepare them for a return to work (Alexander, Baxter, Hughes, & Renda, 2005).

The compulsory activity requirements were tightened in July 2006 upon the introduction of the Welfare to Work reforms, which had greater emphasis on job entry, and a reduction in the age of the youngest child (to 8 years of age) at which PPS recipients had to undertake activity requirements (Centrelink, 2005). In addition, those commencing PPS receipt after July 2006 were no longer eligible for the payment once their youngest child turned eight years old and were instead transferred onto Newstart Allowance (NSA), which is paid at a lower rate and is withdrawn more sharply with respect to the earning of any additional income compared to PPS. This accentuated the expectation that these recipients should be seeking work, and gave them access to the employment-related supports available to NSA recipients, such as job-seeker programs and Career Information Centres. Lone mothers participating in job-focused education or training also have access to supplementary payments; the Pensioner Education Supplement and Training Supplement for PPS recipients and the latter payment and an Education Entry Payment for NSA recipients.

It is interesting to note that between 2005 and 2008 there were large increases in the employment rate of lone parents with school-aged children (as shown in Figure 3), possibly reflecting the impact of the September 2003 and July 2006 reforms. However, employment rates also rose for those with preschool-aged children, who were not directly affected by the policy changes, suggesting that the increases in employment rates for mothers may also be a response to labour market conditions and not just a result of the reforms. In support of this, the unemployment rate for women dropped from 6.2% to 4.5% between 2002 and 2008 (ABS, 2010).

Couple mothers' entitlement to income support depends on their partners' as well as their own income. Those whose partners have low incomes or are income support recipients may be eligible for Parenting Payment Partnered (PPP). This payment is equivalent to PPS, although it is paid at a lower rate. The changes made to eligibility for PPS also affected PPP recipients, although for PPP recipients the new activity requirements and transfer to Newstart Allowance come into effect when the youngest children are six years rather than eight years old. As the recipients were already on a lower payment, the transfer to Newstart Allowance for those commencing PPP receipt after July 2006 did not affect the level of payment received.

In addition to income support for low-income households, families with children can receive Family Tax Benefit A to help with the costs of raising children, with higher rates of assistance for lower income households. Family Tax Benefit B, prior to July 2008, was also available to all lone parents and to couple parents with low personal incomes (including those with higher earning partners). An income test for this payment has subsequently been introduced to exclude higher income families. Other forms of assistance are also available to low-income or income support families, including transport and pharmaceutical concessions and housing assistance.

As the amount of government support available to potential recipients is dependent upon income, there are complex relationships between earned income and eligibility for payments or other forms of assistance. A consequence of this, for some mothers, is low additional returns from paid work relative to income support - once taxation, loss of benefits and costs of paid work (including child care costs) are taken into account. Over recent years, changes to policy have attempted to address this. For example, in 2000, the taper rates for PPS receipt were reduced, allowing recipients to earn more than in the past while retaining benefits (Gregory, Klug, & Thappa, 2008). Other policies include tax offsets and changes in financial assistance for child care. Even so, the financial issues for lone parents moving into employment remain considerable (Gregory et al., 2008; Harding, et al., 2005; Whiteford, 2001).

Looking internationally, several countries have introduced policy measures aimed at increasing lone parents' employment rates, with specific policy goals and approaches varying from country to country. Policy measures include the introduction of programs designed to help parents into employment, enhancement of financial incentives to work, and provision of advice, support and retraining services. They also extend to broader work-family measures, such as the expansion of child care services. For more detailed analyses of these policies in the international context, refer in particular to Adema and Whiteford (2007).

In Australia and in other countries, policy approaches are complicated because care provided by parents is recognised as an important and valuable activity, especially when children are young. For example, in Australia, work requirements for income support recipients do not take effect until children reach school age, which allows these parents to provide full-time care to younger children if they prefer to do so. Other countries offer financial support to parents of young children, regardless of their employment or relationship status (Millar & Evans, 2003).

1.3 Employment transitions: Literature review

Welfare-to-work

Employment transitions have commonly been studied in relation to income support (or welfare) receipt, particularly for lone mothers (e.g., Cancian et al., 1999; Evans, Eyre, Millar, & Sarre, 2003; Gregory, 2002). Interest in this area has expanded since the policy changes in Australia, the New Deal for Lone Parents scheme in the UK, and the Personal Responsibility and Work Opportunity Reconciliation Act 1996 in the US, which all require or encourage lone mothers in receipt of welfare payments to enter paid employment. A particular area of concern is that a large proportion of lone mothers leaving income support are entering jobs that are low paid and insecure and are thus frequently moving back onto income support after short periods in employment (Gregory, 2002).

A UK study found that, while 41% of participants in the New Deal for Lone Parents³ scheme between 1998 and 2002 left income support for paid work, 29% of those who found work returned to income support within a year (Evans et al., 2003). Similar patterns have been found in the US, with roughly 27% of recipients of Temporary Assistance for Needy Families⁴ returning to income support within a year of leaving the program (Acs, Loprest & Roberts, 2001). US research also suggests that a large proportion of lone mothers who left welfare and began working after the 1996 welfare reforms entered temporary jobs (Cancian et al., 1999).

A key concern related to the "churning" of lone mothers between income support and low-wage or low-status jobs is that the financial rewards for working may not make it worthwhile to be in employment, given that lone parents are often entering jobs with low earnings. Some lone parents may even leave employment because they find the costs of working, along with the loss of government support, mean they are not financially better off while working (see studies cited in Richardson & Miller-Lewis, 2002).

These welfare-to-work issues are also relevant to some partnered mothers.

Employment transitions around the birth of a child

Employment transitions of mothers have also been explored in relation to the period of time around the birth of a child, although comparisons between lone and couple mothers appear limited. Joesch (1997) and Ondrich, Spiess, Yang, and Wagner (2003) included relationship status in their analyses but found no differences in the timing of return to work according to this variable. International findings with regard to differences between lone and couple mothers may not apply to Australia, since in many OECD countries the employment rate of lone mothers is not significantly below that of couple mothers (see Figure 1). For the UK, where the employment rate of lone mothers is relatively low, Dex, Joshi, Macran, and McCulloch (1998) found that lone mothers actually had a marginally faster return to work after a birth compared to mothers with an employed partner. In Australia, Baxter (2008) found that lone and couple mothers were not significantly different in the timing of their return to work when their pre-birth employment status was also taken into account. However, lone mothers did have a lower likelihood of returning to work by 12 or 18 months after a birth when pre-birth employment status was not included in the analysis.

Mothers' employment transitions are likely to vary with children's age, with an expectation that not-employed mothers might be more likely to enter work as children get older (resulting in higher employment rates for mothers with older children; see Figure 3). This is consistent with Barrett's (2002) analyses of transitions off income support in Australia. He found that exits off payment were lowest for those with a child aged under 5 years, and increased for those with older children.

Relationship change and employment

As lone parenthood in Australia largely comes about due to relationship breakdown rather than from death of a partner or births to single women (ABS, 2003; McHugh & Millar, 1996), it is relevant to consider the event of separation as a factor for lone mothers in explaining their employment rates. In a study of separated and divorced Australian mothers of children aged less than 18, Walters (2002a) found that employment rates dropped in the first year after the marriage ended (from 63% to 55%) and did not return to pre-separation employment rates until five years later. Similarly, in a UK study, Paull (2007) found that mothers' reduced employment rates after relationship breakdown were due to a combination of high rates of leaving employment after separation and low rates of entering employment in the months following. Funder (1993) found similar results and also identified occupation, level of education and early re-partnering as additional strong predictors of post-separation employment. Walters also found declines in work hours post-separation, with reductions in full-time employment rates and increases in part-time employment rates occurring immediately after separation and in the year following.

The other possibility for separated women is that their relationship breakdown might result in a financial need to work, which may even be apparent in the lead up to separation (Johnson & Skinner, 1986). Jeon (2008) and Johnson and Skinner (1986) found that recently divorced non-working women (not just mothers) were more likely to enter employment than other non-working women, which they viewed as a response to income reduction. Jeon also found that women who had recently married were more likely than others to leave employment, though the relationship was not as strong. Whether these effects would be apparent for mothers, rather than all women, is not clear.

Labour market factors

Mothers' labour market transitions are likely to be a factor of both their own characteristics as well as the characteristics of the labour market and the nature of the employment they are able to undertake.

Labour market transitions are likely to differ according to whether mothers are employed in or seeking casual versus permanent employment, and the extent to which they are able to attain employment in jobs of higher status and earnings. Buddelmeyer, Wooden & Ghantous (2006) analysed casual employment and showed that men and women who are employed in casual jobs at one point in time are more likely to be out of work at a later time than those employed in non-casual jobs. Similarly, Perkins and Scutella (2008) showed that those employed in low-paid jobs are less likely than other employed persons to be employed at a later time.

Levels of human capital are an important factor here, as mothers with higher levels of human capital, measured in terms of education or work experience, are more competitive in the labour market, and therefore able to find employment in jobs with better paid and/or working conditions. Lone mothers are at a disadvantage in this respect, given their relatively low levels of human capital (Harding et al., 2005). This is evidenced by the finding that employed lone mothers of young children are more likely to be employed in casual jobs and jobs of lower occupational status, and less likely to have access to family-friendly work arrangements than couple mothers (Baxter, Gray, Alexander, Strazdins, & Bittman, 2007).

These findings indicate that lone mothers are likely to face disproportionately higher rates of movement out of employment. However, if turnover in these jobs is relatively high, we can perhaps also expect that mothers seeking lower status or casual jobs may be able to move into these jobs more easily than mothers seeking higher status, permanent positions, for which turnover is lower. Of particular importance is whether higher turnover in casual or low-paid jobs is driven by employees or by employers. It is plausible that mothers with lower levels of human capital may be at greater risk of being subject to employer-driven turnover, and therefore being less able to leave or take up employment at a time that matches their needs.

Movements into and out of employment are strongly associated with past employment experience. Those who have spent more time in employment are more likely to remain employed if already employed, or to enter employment if not employed. In the economics literature, this is examined in relation to "duration dependence" or "state dependence", which reflects that people are much more likely to stay in their state of employment than they are to change it (Buddelmeyer et al., 2006; Haynes, Western, Yu, & Speliak, 2008; Knights et al., 2000; Stromback et al., 1998). This may be particularly important for this paper, if lone mothers and couple mothers have quite different employment histories. This might come about if, from the time of relationship separation, lone mothers have low employment rates, or if those who become lone mothers (through a birth or separation) have relatively low employment rates even before lone motherhood.

Work-family balance, work orientation and child care

In discussing the employment rates of mothers, it is important to consider that mothers often prefer to look after children while they are young, and thus take a break from employment at this life cycle stage. Therefore, a significant proportion of not-employed lone and couple mothers actually prefer not to work (McHugh & Millar, 1996). Whether lone and couple mothers differ in this regard is not clear.

Differences in eligibility requirements for income support payments may mean lone and couple mothers do not have the same incentives and disincentives to look for work. Lone mothers, however, have more barriers to employment; for example, they have relatively high rates of physical and mental health problems (Butterworth, 2003).

Another factor that may influence mothers entering and remaining in employment concerns the difficulties they experience in reconciling their work and family responsibilities. In exploring the differences in work-family strain of lone and couple mothers, Baxter and Alexander (2008) found that employed lone mothers were more likely than couple mothers to agree that they had missed out on home or family activities and more likely to agree that their family time was less enjoyable and more pressured because of work responsibilities. These differences, however, appeared to be related to the different characteristics of lone and couple mothers, rather than being a factor of lone parenthood itself. Nevertheless, lone mothers are more likely to be constrained in their child care options, given that they cannot readily share the care of the child with another resident parent, and this may place barriers on their ability to enter employment, and cause difficulties in sustaining employment (McHugh & Millar, 1996).

The cost and availability of child care may also be a preventative factor for mothers wanting to enter and stay in paid work. Past studies have found that the more affordable and available child care is, the more likely it is that mothers will be in paid work or vice versa (Avdeyeva, 2006; Blau & Ferber, 1992; Breunig & Gong, 2010; Gong, Bruenig & King, 2010; Gornick & Meyers, 2003). Child care costs may be more of an issue for lone than couple mothers, as lone mothers are more likely to use formal child care, such as child care centres, rather than informal child care, and couple mothers are more likely to use informal child care, particularly grandparent care (ABS, 2009).

1.4 Summary

The likelihood of leaving or entering employment is expected to vary with factors such as the availability of suitable jobs, the ability to reconcile work and family responsibilities and the potential financial gain from employment. Lone and couple mothers are likely to differ in relation to these factors and their impact on employment entry and exit. The analyses in this paper do not attempt to isolate which of these factors is most influential, but by taking into account a range of characteristics (work histories, education, numbers and ages of children, health status and age) in addition to lone parenthood, we can provide a clearer picture of whether this set of factors affects lone mothers' employment transitions more than those of couple mothers.

2. Data and methods

This paper uses multiple waves of the HILDA calendar data to analyse lone and couple mothers' employment transitions. Throughout this paper, couple mothers include married as well as cohabiting mothers, and both lone and couple mothers are mothers with dependent children aged under 15 years old.

In this section we describe the HILDA survey, the calendar data, the derivation of relationship status and other data items and the methods used.

2.1 The Household, Income and Labour Dynamics in Australia survey

HILDA is a nationally representative, annual panel survey, which commenced with Wave 1 in 2001. The sampling unit for the survey is households, with information being gathered on each member of the sampled households and interviews conducted with household members aged over 15 years. For Wave 1, 11,693 households were sampled from 488 areas (Census Collection Districts) across Australia. Members of 7,682 households completed interviews, resulting in 13,969 completed individual interviews and a response rate of 66%. While sample size has declined over the waves due to attrition, at each wave new members to households are added in. For more information about HILDA, refer to Watson and Wooden (2002a).

This paper uses data from the first eight waves of HILDA, primarily using the employment calendar component of the survey. However, Wave 1 data have been excluded for much of the analyses, primarily because significant changes to this component of the survey were made after Wave 1 in order to correct some design issues that affected data collection (Watson & Wooden, 2002b). Another contributing factor was that Wave 1 lacked key information about the recent history of relationships, so it was not possible to accurately match relationship status to the calendar data collected in this wave.

All waves were used for a brief descriptive analysis of the characteristics of employed and not-employed mothers by relationship status.

The analyses were limited to mothers of children aged under 15 years, which was around 2,000 persons at each wave. The analyses of employment transitions are based on unweighted data.

2.2 Calendar data

The calendar component of HILDA is obtained at each wave of the survey, with details of work and study activities being collected for the period starting from 1 July of the previous year up to the survey date.⁵ In relation to work, respondents are asked to indicate how many jobs they have had over this period and to identify the dates within which they worked in each of those jobs. From this information, a series of job indicator variables are created that identify whether, at the start, middle and end of each month covered by the calendar, each respondent worked in each job.

These calendar data were aggregated for analysis in this paper. The job-level information was collapsed to obtain an indicator of whether or not respondents were employed in any job in any particular period. Note that being "employed" includes periods of paid leave, and therefore includes paid maternity or paid holiday leave. Those classified as "not employed" include the unemployed as well as those not in the labour force. Table 1 shows a summary of these data, just using Wave 2 of HILDA, for mothers with children aged under 15 years.

Note: ^a Data for total mothers, for all waves, is given in Appendix 2 (Table A2).

At the time of the Wave 2 survey, couple mothers were more likely to be employed (62%) than lone mothers (51%). Mothers (both lone and couple) employed at Wave 2 reported having spent, on average, 91% of the previous financial year employed.

Looking at the financial year data for all mothers, 55% of couple mothers and 43% of lone mothers were employed for the whole of the previous financial year. The proportion spending some time, but not all year, in employment was 14%, with little difference between lone and couple mothers. The balance, 30% of couple mothers and 42% of lone mothers, spent no time in employment. These data, along with those for all years in Table A2, indicate that the majority of mothers are not transitioning between employment and non-employment over a one-year period, and also indicate that the rate of transition may not vary between couple and lone mothers.

To explore employment transitions over a more extended period, each respondent's calendar data from each wave was matched to develop a complete calendar across all waves for each individual (that is, from 1 July 2000 to the respondent's interview date in 2008). Note that while respondents were surveyed every 12 months, calendar data for each wave covered a 14-18 month period, therefore creating an "overlapping seam", or a period of time in which data were available from two waves. In these analyses, the job spell information reported closest to the period being recalled was taken as the more accurate, based on prior analyses of these data (Watson, 2009). If the more recent of the two dates had a missing value attached, the alternative source was used. More information about the overlapping seam is given in Appendix 1.

The calendar data were used to capture movements into and out of employment in consecutive time periods. First, the three within-month time periods (start, middle and end) were aggregated to one month, such that respondents were classified as being employed in a month if employed in any of those three time periods. This was to simplify the analyses and to allow a better match with data covering respondents' characteristics, which were not collected at the within-month level. Changes in relationship status, our key characteristic, could be matched well to the monthly level (discussed further below). However, by collapsing the data to monthly periods, some employment transitions were missed, as respondents who were classified as employed may not have been employed for the whole month. Analyses presented in Appendix Table A1 show that the number of cases affected by this is relatively small, resulting in the same overall relationships being observed, regardless of which time period is used. Respondents who completed one or more waves are included in this analysis, and if employment data were missing for a wave, due to non-response or other reasons, the employment status was set to missing just for those periods affected.

Data were then converted to a person-month format, so that for every person, for every calendar month (that is, year by month), there is an indicator of whether or not they were employed at that time. The resultant dataset is therefore, by design, the number of respondents times the number of calendar months for which employment data were reported. As a result, there are significantly more records to analyse than there are respondents, as all respondents have multiple monthly records. For example, Table 2 shows that for employment transitions during 2002, there were over 17,500 person-months for 1,632 couple mothers and over 4,600 person-months for 517 lone mothers. Across all years, there is a total of almost 114,000 person-months for couple mothers and over 32,000 person-months for lone mothers.

Note: The person numbers refer to the total number who were couple mothers and the total number of who were lone mothers at some time within a calendar year (or across any year). This double-counts respondents, since across all waves 448 women changed between being single and couple mothers between waves. The total numbers of persons only count each mother once and are different to the totals presented in Table 1, as these are the number contributing data to each calendar year, rather than the number at a particular wave. Only those with non-missing employment status data are included.

Employment data in consecutive months were used to derive measures of employment transitions, specifically to derive whether respondents stayed employed, stayed not-employed, left employment, or entered employment (see Table 4, Section 3). For each month, two key measures were derived: for those employed in one month, the likelihood of leaving employment the next month; and, for those not employed in one month, the likelihood of entering employment the next month (see Table 5, Section 3). This paper focuses on these particular measures, comparing couple and lone mothers, along with other characteristics that are likely to explain such employment transitions.

2.3 Relationship status

As noted previously, the sample at each wave was restricted to mothers with dependent children aged up to 15 years old. Mothers' relationship status was then used to differentiate between lone and couple mothers, with couples including those in married and de facto (or cohabiting) relationships. Relationship status was available at each wave, and relationship changes that occurred between waves were identified using the detailed questions on the timing of any relationship commencement or separation. This enabled relationship status to be derived at the monthly level. See Appendix 1 for more information.

Most mothers in the sample did not change relationship status across waves, although lone mothers were more likely to have experienced a change in relationship status at some time. For example, of the mothers who were partnered at the time of the Wave 8 interview, only 8% had been single at some point over the previous seven waves. For those who were lone mothers at Wave 8, 54% had been in a cohabiting relationship or marriage at some time between Waves 1 and 8. The remaining 46% of lone mothers at Wave 8 had been lone parents for the whole survey period, including those who entered the sample through the birth of a child between Waves 1 and 8, although 35% were in a couple relationship prior to Wave 1.

Relationship status is included in these analyses as a time-varying characteristic, and we use relationship status in one month to calculate likely exits or entrances into employment in the following month.⁶

2.4 Other characteristics

While relationship status was the main variable of interest, other characteristics were explored, given their established relationship with maternal employment rates and transitions (e.g., Gray et al., 2006). Specifically, we include measures of ages and numbers of children, educational attainment, health status, country of birth and age of mother, and mothers' recent work experience. See Box 1 for more information.

The differences in characteristics of lone and couple mothers are apparent in Table 3, with lone mothers having spent less of the prior year in employment and having lower levels of education and poorer self-reported health. Lone mothers are also more likely than couple mothers to have older children (and less likely to have children under 3 years old) and are more likely to only have one child. Differences between lone and couple mothers' characteristics are still apparent when employment status is taken into account.

Note: Percentages are calculated based on the distribution of these variables over the pooled person-month data across Waves 2-8.

2.5 Methods

In order to analyse employment transitions of lone and couple mothers, we first present in Section 3 an overview of the employment data and transitions. The two types of employment transitions - leaving employment for employed mothers, and entering employment for not-employed mothers - are then analysed separately. Each of these sets of analyses involves: (a) a discussion of the employment or labour market characteristics of those "at risk" of making that employment transition; (b) descriptive analyses of how employment transitions vary by relationship status, as well as the other characteristics identified in Box 1; and (c) multivariate analyses of the employment transition. More detail about each of these steps is given below.

Employment and labour market characteristics of employed and not-employed mothers

A range of employment or labour market characteristics were selected for examination from each wave of HILDA to help identify how lone and couple mothers may differ in their likelihood of entering or leaving employment. Characteristics explored for employed mothers were: job contract (whether employed permanently, casually or fixed term), job tenure (number of years in current job), hours worked, gross wages, occupational status (measured on scale of 1 to 100), job satisfaction (measured on a 10-point scale), whether or not they left a job in the last year and number of jobs in the last year. Not-employed mothers were examined in relation to whether they were unemployed or not in the labour force, and the hours and income they would accept if they became employed. For mothers not in the labour force, reasons for not looking for work and whether or not they wanted a job were examined.

These analyses provide some insights as to why lone and couple mothers may have different outcomes in the labour market. The detailed employment or labour market characteristics were not, however, included in the analyses of monthly employment transitions, since they were not available at the monthly level.

Descriptive analyses of employment transitions

Presentation of the employment transitions results begins with descriptive analyses, whereby employment transitions are compared for lone and couple mothers, by each of the explanatory variables one at a time. The analyses of leaving employment, presented first, gives the percentage of mothers leaving employment in the next month, given they were employed in a given month. Similarly, in the analyses of entering employment, the percentages are based on those who were not employed in a given month, and show the transitions that occurred from that month to the next. Each figure shows the percentage making the transition, with the 95% confidence interval also indicated.

Multivariate analyses of employment transitions

The multivariate analyses allow us to ascertain whether relationships between lone parenthood and employment transitions are apparent once the other selected characteristics of mothers are also taken into account. Further, they allow for examination of how much these characteristics explain differences in employment transitions. This is important given that lone and couple mothers vary considerably in relation to these factors.

As with the descriptive analyses, the multivariate analyses were conducted separately for each type of transition.

To do these analyses, the data from all months were pooled, and the analyses estimated the likelihood of leaving/entering employment at time t + 1, for those who were employed/not employed at time t. The models included explanatory variables, measured at time t, as well as variables to control for the year and quarter to which the data refer.

As the outcome variables are binary (either leaving or entering employment), logistic regression is an appropriate method of analysis. However, as there are multiple records (that is, months) per person, more sophisticated techniques are required to allow for the within-person correlations. This is important in this type of analysis, as this person-level variation could be quite high, especially considering that the models include only a fairly small set of person-level explanatory variables. Therefore, random effects (RE) models were used as these models allow us to determine how employment varies with lone parenthood, and with the range of background characteristics. The coefficients in these models estimate the difference in the log odds of being employed associated with the presence of a particular characteristic, such as being a lone parent as opposed to a couple parent. Because the estimation is based on multiple records per person, the coefficients represent both differences across respondents (at any one month) and differences within respondents (across months). Some characteristics do not change at all across time (e.g., country of birth and, usually, education), while some change considerably (e.g., age and ages of children) or have the potential to change (e.g., relationship status). For those variables that may change across months, the estimated coefficient will reflect these changes among individuals as well as between individuals. This is the case for the relationship status variable, which for some mothers will remain constant across the waves, while for others will capture changes due to separation or partnering. These coefficients cannot be used to draw conclusions about causal relationships, but instead are used to describe associations between variables.

In presenting the results, marginal effects were calculated, as these estimates are easier to interpret than the model coefficients. The marginal effect is the change in the probability of leaving or entering employment associated with a particular characteristic (e.g., lone parenthood), holding constant the value of all other explanatory variables. For binary variables, the marginal effect is for a change in the value of the variable from zero to one; that is, from not having the characteristic to having it, as is the case with lone parenthood. For categorical variables, such as highest level of education, the marginal effect is relative to that variable's reference category, which is indicated in the presentation of results. Marginal effects were calculated assuming a random effect of zero (which is required when calculating marginal effects after RE estimation), and with other variables set at their sample mean.

Models were also estimated separately for lone and couple mothers, to determine whether any of the factors controlled for had a stronger (or weaker) effect on lone rather than couple mothers. To test whether there were significant differences in the coefficients for lone and couple mothers, models were also estimated in which relationship status was interacted with each of the variables. These analyses revealed no statistically significant differences (at the p < .05 level) between the two models' coefficients, for either leaving or entering employment.

The multivariate approach used here is far simpler than those of other studies (e.g., Buddelmeyer et al., 2006; Haynes et al., 2008; Knights et al., 2000). This is because its primary purpose is to examine how transitions vary with relationship status, after taking into account other variables. More sophisticated approaches can be explored in future analyses of these data. Also, it is possible to extend these analyses to consider whether effects vary for different groups of mothers (e.g., lower educated lone mothers versus higher educated lone mothers; lone mothers with younger children versus those with older children), although this is beyond the scope of this paper, since such analyses are not necessary to address our main question of interest - whether lone and couple mothers' employment transitions differ. Also, the sample size would limit opportunities for further breakdown.

3. Results

3.1 Overview of HILDA employment and transitions data

This first subsection provides an overview of the employment data, and an introduction to the employment transitions analysed in more detail in subsequent sections.

Figure 4 shows how employment varies with relationship status and age of youngest child over time, using the monthly calendar data derived from HILDA. To be comparable with the ABS data used in Figure 3, June data from HILDA have been presented. While the employment rates in Figure 4 are somewhat higher than those presented in Figure 3, the overall patterns are the same.⁷ Across all years, lone mothers were less likely to be employed than couple mothers, regardless of the age of their youngest child, although the employment rate of lone mothers has increased gradually since 2004.

Figure 4 Employment rates of mothers with children aged under 15 years, by relationship status and age of youngest child, June 1984 to June 2008 (HILDA)

Source: HILDA, Waves 2-, pooled monthly calendar data

Looking at the data as monthly employment transitions instead, Table 4 shows that, consistent with Table 1, the proportion moving between employment and non-employment is small relative to the proportions remaining employed or not employed. Ninety-eight per cent of mothers did not change employment state from one month to the next, with 66% staying employed and 32% staying out of employment. Lone mothers were more likely than couple mothers to stay not employed (41% compared to 30%)

Note: Percentages do not total 100% due to rounding.

The low proportions transiting into or out of employment is not surprising - various other studies of employment transitions have shown that people are much more likely to stay in their state of employment than they are to change it (Buddelmeyer et al., 2006; Haynes et al., 2008; Stromback et al., 1998). Table 5 shows that of those employed in any one monthly period, only 1.5% are not employed in the next month, with a higher exit rate for lone mothers (2.0%) compared to couple mothers (1.3%). Of those who are not employed in one monthly period, a slightly higher percentage moves into employment (3.6%), with no difference between lone and couple mothers.

3.2 Leaving employment: Employment transitions of employed mothers

Characteristics of employed mothers: Those "at risk" of leaving employment

Do the employed lone and couple mothers have jobs with different characteristics, such that we would expect different rates of leaving employment according to relationship status? For example, as reported previously, those in casual or low-paid jobs are more likely to leave employment than those with other types of job contract or higher wages (Buddelmeyer et al., 2006; Perkins & Scutella, 2008).

Using the HILDA survey data at each wave, Appendix Table A3 provides a summary of some of the characteristics of, and perceptions about, jobs held by lone and couple mothers. Here we just comment on the two factors that were most different for lone and couple mothers: job contract and job tenure.

Looking first at job contract, while similar proportions were in permanent employment, lone mothers were more likely to be classified as casual, contract or fixed-term. On the other hand, couple mothers were more likely to be employed in their own business. For example, in 2008, among employed mothers, 55% of lone and 54% of couple mothers were in permanent jobs; 38% of lone and 29% of couple mothers were in casual, contract or fixed-term jobs; and 7% of lone and 16% of couple mothers were employed in their own business. This higher rate of casual work of lone mothers may be associated with their having a higher rate of exit from work than couple mothers.

Looking at job tenure, among employed mothers, lone mothers were more likely than couple mothers to have started their current job in the preceding 12 months (in 2008, 30% of lone and 18% of couple mothers). Similarly, lone mothers were more likely to have left their job between waves (21% of lone mothers and 10% of couple mothers in 2008). This may be a result of job churning rather than employment exit; however, there was no difference in the average number of jobs lone and couple mothers had held in the last financial year. The higher incidence of casual work among lone mothers may suggest some instability in their employment.

Other characteristics, examined in Appendix Table A2, were in some ways quite similar for lone and couple mothers - including their hours of work (lone mothers working slightly longer hours in most years) and gross wages (lone mothers earning slightly less). Overall, job satisfaction (as a mean score) was similar for lone and couple mothers.

Employed mothers leaving employment: Descriptive analyses

This subsection focuses on the transition out of employment, presenting differences for lone and couple mothers, first examining changes over the years, and then according to the range of explanatory variables outlined in Box 1.

Table 4 showed that lone mothers were more likely to leave employment, if employed in the previous month, than partnered mothers. This is also shown, in Figure 5, to be true across all years, although the differences were more apparent in some years than in others (in some years, such as 2008, the difference between lone and couple mothers was not significant at the 95% confidence level).

Figure 5: Percentage leaving employment, by relationship status, mothers employed in the previous month, 2002-08

Figure 6 shows that there is a very strong relationship between recent work history and the likelihood of leaving employment. There were very low transition rates for those who were employed for most or all of the previous financial year, but much larger transition rates for those who were employed for a lesser amount of time or not employed at all. In each of these groups, differences between lone and couple mothers are apparent, although they were not statistically significant for those with little or no recent employment experience.

Figure 6: Percentage leaving employment, by relationship status and recent work history, mothers employed in the previous month

Looking at differences by level of education, overall, higher levels of educational attainment are associated with a lower likelihood of leaving employment (Figure 7). Comparing relationship status within each education level, no differences are apparent between higher educated lone and couple mothers. However, for those with lower levels of education, lone mothers had a higher rate of exit from work than couple mothers, with the widest gap being for those with incomplete secondary education.

Figure 7: Percentage leaving employment, by relationship status and highest level of education, mothers employed in the previous month

Employed women are likely to leave employment before the birth of a child, even if only for a short time to recover from childbirth and to care for a newborn. We would therefore expect relatively high rates of leaving employment around the birth of a child, often just beforehand. However, because these data included only those parents who already had children, employment exits that occurred prior to the first birth are not captured. Also, as noted previously (see Section 2.2), periods of paid leave from employment are counted as employment. As such, mothers who remain employed because they are on a paid absence from work after a birth (e.g., paid maternity leave or holiday leave) are not captured as an exit from employment. However, as unpaid maternity leave should be counted as "not employed", moving from a period of paid to unpaid maternity leave should be captured as an exit from employment.

Figure 8 shows that for both lone and couple mothers, the likelihood of leaving employment was considerably higher for those with the youngest children, which may be related to movements out of paid leave. It may also indicate that mothers more often take on short-term or casual jobs when they have very young children. Lone mothers were more likely to leave employment than couple mothers over all these child ages, but the gap was larger for those with very young children.

Figure 8: Percentage leaving employment by relationship status and age of youngest child, mothers employed in the previous month

The number of children, on the other hand, did not make a significant difference to the likelihood of leaving employment (Figure 9), although for families with 3 or more children, lone mothers were much more likely to exit employment than couple mothers. Lone mothers were also more likely to exit employment than couple mothers in one- and two-child families, but the differences between these groups of mothers were much smaller than in the larger families.

Figure 9: Percentage leaving employment by relationship status and number of children, mothers employed in the previous month

For couple mothers, those with poor self-reported health had a higher likelihood of leaving employment, compared to those with good health. For lone mothers, however, these differences according to health status were not statistically significant (Figure 10).

Figure 10: Percentage leaving employment by relationship status and self-reported health status, mothers employed in the previous month

Country of birth had little overall association with mothers' employment exit (Figure 11). In regard to relationship status, for those born in Australia, lone mothers were more likely to exit employment than couple mothers; however, there was no difference between lone and couple mothers for those born overseas.

Figure 11: Percentage leaving employment by relationship status and country of birth, mothers employed in the previous month

Employed mothers leaving employment: Multivariate results

Multivariate analyses were used to explore these relationships further, to determine which factors remained significantly associated with a higher probability of exiting employment, when other variables were taken into account. The variables included in the analyses are those shown in Table 6, including the variables explored above, as well as the year and quarter of the year, which capture seasonality and changes over time. Job characteristics were not included as these data were not available at the monthly level.

Note: Random effects model also included controls for missing health information and missing employment experience information. a Rho measures proportion of total variance contributed by person-level variance. *** p < .001, ** p < .05, * p < .01.

After taking into account all the variables, a significant difference between lone and couple mothers remained evident. Table 6 shows that when other characteristics are taken into account and held at their mean values, lone parenthood was associated with a marginal effect of 0.4%. This means that the rate of exit from employment was predicted to be higher for lone than for couple mothers by 0.4%. While this difference is small, it is noteworthy when considered in relation to the overall transition rates for lone and couple mothers, which were themselves quite small (2.0% and 1.3% per month respectively).

These results show that the higher probability of leaving employment for lone mothers was not entirely related to differences between lone and couple mothers in the characteristics explored. As we did not include information about the jobs held by lone and couple mothers in these analyses (since this information was not available on a month-by-month basis), some of these differences may be associated with lone mothers' higher rates of casual employment and shorter job tenure.

The multivariate analyses also confirm:

• Transitions out of employment are most likely for those with little or no recent work history, while those with moderate amounts of work history are also more likely to transition out than are those who were employed for most or all of the previous year. The marginal effects show that compared to those who were not employed or were employed for less than 20% of the previous year, those employed for 80-100% of the previous year were predicted to have a 1.1% lower chance of leaving employment and those employed 20 to under 80% of the previous year were predicted to have a 0.8% lower chance of leaving employment.
• More highly educated mothers are less likely to leave employment. Using the marginal effects, compared to those with a bachelors degree or higher, those with complete secondary/certificate/diploma were predicted to have an increased likelihood of leaving employment of 0.2%, and those with incomplete secondary schooling were predicted to have an increased likelihood of employment exit of 0.4%.
• Exits from employment are more likely for mothers with younger children. Compared to those with their youngest child aged 0-2 years, marginal effects show that those with their youngest child aged 10-15 years were predicted to have a 1% lower chance of leaving employment. Those with youngest children in the middle two age groups also had lower predicted chances of leaving employment than those with the youngest children (0.5% for those with youngest children aged 3-5 and 0.7% for those with youngest children aged 6-9).
• Some associations were also apparent for mothers' age and health status, while the number of children and country of birth were not significantly related to the likelihood of exiting from employment. The controls for the year and quarter of year showed that mothers were more likely to exit employment in the third or fourth quarters of the year compared to the first, but there were no significant findings in relation to the whole year.

These models were also estimated separately for lone and couple mothers, to assess whether any factors had a stronger effect for one relationship group over the other, as seemed to be the case in some of the descriptive findings above. Results are presented in Appendix Table A5. Comparison of each coefficient in the lone mother model to the equivalent one in the couple model revealed that none of the coefficients were significantly different from each other. That is, there is no evidence that in the presence of a certain characteristic (such as low education), lone mothers behave differently to couple mothers in terms of employment exit.

Putting the multivariate results together with the different characteristics of lone and couple mothers (Table 3), lone mothers' rate of exit from employment is likely to be greater than that of couple mothers because lone mothers tended to be younger, had lower levels of education, poorer self-reported health and lower levels of employment experience. On the other hand, the opposite is true with regard to the ages of their children, as lone mothers were less likely than couple mothers to have very young children, and exit rates from employment were higher when children are younger.

3.3 Entering employment: Employment transitions of not-employed mothers

Characteristics of not-employed mothers: Those "at risk" of entering employment

This subsection examines characteristics of mothers who are not employed, to see whether barriers to entering work may be different for lone and couple mothers. Appendix Table A4 shows a range of relevant indicators. Here we focus on those that relate specifically to job-seeking behaviours and barriers to the take-up of employment.

Based on information collected at the time of the survey, the majority of not-employed lone and couple mothers were not in the labour force, as opposed to being unemployed, indicating that the majority of not-employed mothers were not actively engaging in job search. However, lone mothers were more likely than couple mothers to be unemployed, which does suggest they were more likely to be actively seeking work. For example, of not-employed mothers in 2008, 15% of lone mothers and 6% of couple mothers were unemployed. Also, substantially more lone than couple mothers who were not in the labour force said they wanted a job (for example, 44% of lone mothers and 24% of couple mothers not in the labour force in 2008). Furthermore, in all years, lone mothers were more likely than couple mothers to say they could start work in the next four weeks (e.g., 31% of partnered mothers and 49% of lone mothers not employed in 2007). While these data suggest that lone mothers may have more opportunity to move into employment, their characteristics - such as lower levels of education, poorer health and weaker attachment to employment (as indicated by recent work history) - may mean that despite their more active job search, they may not necessarily move into employment at a higher rate than couple mothers. It is also possible that lone mothers may be more particular about their requirements for work to be at specific times, or to allow certain flexibilities that couple mothers could manage without.

The majority of mothers who were not in the labour force said they were not looking for work because they prefer to look after their children (about two-thirds). Couple mothers were a little more likely than lone mothers to give this reason for not looking for work, in all years except 2008. Lone mothers appear to have more difficulties finding child care (e.g., in 2008, 7% of couple mothers not in the labour force, compared to 10% of lone mothers, with greater differences evident in all other years). Lone mothers also appear to be affected more than couple mothers by their own illness, injury or disability (e.g., 10% of couple mothers not in the labour force, compared to 18% of lone mothers in 2008), consistent with the data presented in Table 3. Lone mothers, therefore, may be more likely to be affected by particular barriers (ill health, access to child care) than couple mothers, potentially adversely affecting their rate of transition into employment.

These data provide a mixed picture, with lone mothers expressing more interest in being in employment, but at the same time, appearing to have more barriers to entering employment.

Not-employed mothers entering employment: Descriptive analyses

This subsection now turns to the likelihood of entering employment for those who are not employed, again making comparisons according to relationship status and other factors as abovementioned. In earlier results (Table 4), differences between not-employed lone and couple mothers in their likelihood of entering employment were not apparent. Figure 12 also shows no significant difference for any years, although during 2005, lone mothers' rate of moving into employment was higher than it was in earlier years.

Figure 12: Percentage entering employment by relationship status, mothers not employed in the previous month, 2002-2008

As with the transition out of employment, the transition into employment proved to be strongly associated with recent work history (Figure 13). Significant differences by relationship status were not apparent within any of the categories of recent work history, but overall, those with less employment in the previous year were much less likely to enter employment than those who had been employed for a large part or all of the previous year.

Figure 13: Percentage entering employment by relationship status and recent work history, mothers not employed in the previous month

Level of education has a very strong association with the likelihood of moving into employment (Figure 14), particularly with regard to the distinction between those with and without tertiary qualifications. Those with a bachelors degree or higher had a significantly higher rate of movement into employment compared to those with lower levels of education.

Figure 14: Percentage entering employment by relationship status and highest level of education, mothers not employed in the previous month

Since we know that the employment rates of mothers are very dependent upon the age of their youngest child, we expect to find some relationship between age of youngest child and the rates of entering employment. Figure 15 shows that transitions into employment are least likely for lone mothers when they have a child aged under 3 years old; however, for lone mothers there is little variation in employment entry rates for the other three age groups. For couple mothers, it is those with their youngest child in the second youngest age group (3-5 years) who are least likely to enter work.

Figure 15: Percentage entering employment by relationship status and age of youngest child, mothers not employed in the previous month

The likelihood of moving into employment decreases as the number of children increases (Figure 16), so that for both lone and couple mothers, those with one child were the most likely of all family size groups to enter employment. There were no significant differences within the other family size groups in the rate at which lone and couple mothers moved into employment.

Figure 16: Percentage entering employment by relationship status and number of children, mothers not employed in the previous month

In general, mothers with poorer self-reported health were less likely to enter employment, if not employed; however, there were no significant differences between lone and couple mothers with the same health status (Figure 17).

Figure 17: Percentage entering employment by relationship status and self-reported health status, mothers not employed in the previous month

There was some variation in employment entry by country of birth (Figure 18). Those born outside Australia in an English-speaking country were most likely to enter employment (significant only for couple mothers) and those born in a non-English speaking country were the least likely to enter employment. There were no significant differences between lone and couple mothers within any of these groups.

Figure 18: Country of birth of those entering employment, mothers not employed in the previous month, by relationship status

Not-employed mothers entering employment: Multivariate analyses

As with the analyses of leaving employment, this subsection now explores how transitions into employment vary using multivariate analyses. Table 7 presents the regression coefficients and the associated marginal effects for the model in which lone and couple mothers are combined, and the range of explanatory variables included.

Note: Random effects model also included controls for missing health information and missing employment experience information. 
a Rho measures proportion of total variance contributed by person-level variance. *** p < .001, ** p < .05, * p < .01.

Consistent with the initial analyses of employment transitions in Table 4, the difference between lone and couple mothers in their rate of entering employment was not statistically significant. This is interesting, given the findings presented earlier that lone mothers are more likely to be actively seeking work, and more likely to want a job and be ready to start work. These earlier findings also showed that lone mothers appear to have more barriers that may impede their take-up of employment, and the lack of difference in entrance rates to employment between lone and couple mothers therefore suggests that these barriers do indeed result in lone mothers having no greater take-up of employment, despite their apparently stronger willingness or desire to work.

Several factors were found to explain variation in the rate of entering employment:

Recent work history was very important, with those who spent more of the previous year in employment being more likely to enter employment. Using the marginal effects, compared to those employed for less than 20% of the previous year, those who were employed for 20-79% of the previous year were predicted to have a 2.4% greater chance of entering employment, and those who were employed for 80% or more of the previous year were predicted to have a 2.7% greater chance of entering employment.

More highly educated mothers were more likely to enter employment. Compared to those with a bachelors degree, those with incomplete secondary schooling were predicted to have a 4.8% lesser chance of entering employment and those with complete secondary schooling/diploma/certificate had a 2.7% lesser chance of becoming employed.

Entering employment was least likely for mothers with younger children, and highest for those with children aged 6-9 or 10-15 years. Compared to those with their youngest child aged 0-2 years, those whose youngest child was aged 10-15 years were predicted to have a 2.1% greater likelihood of entering employment. The predicted increase was the same for those whose youngest child was aged 6-9 years. For those with a 3-5 year old youngest child, the predicted increase was lower, at 0.8%.

Number of children, health status and country of birth also had significant associations with the likelihood of entering employment. Further, compared to 2008, mothers were less likely to enter employment in 2002 or 2003. Also, mothers were more likely to enter employment in July-September compared with January-March.

As with the multivariate analyses of leaving employment, separate models were also estimated for lone and couple mothers, and these estimations are given in Appendix Table A6. As with leaving employment, none of the lone mother coefficients were significantly different from those for couple mothers.

Discussion

This paper set out to identify differences between lone and couple mothers in their likelihood of entering and exiting employment, in order to help in understanding the different employment rates of lone and couple mothers. These results show that lone mothers were more likely to leave employment than couple mothers, but they were no less likely to enter employment. These same findings have been observed in analyses of lone mothers' employment in the UK (Evans et al., 2004).

While the higher rate of exiting employment by lone mothers is interesting, the lack of difference between lone and couple mothers in rates of entering employment is also noteworthy because lone mothers seemed to express a greater desire and availability to work than couple mothers, suggesting they were faced with more barriers to finding or gaining suitable jobs.

These analyses showed that quite small percentages of mothers, whether lone or couple, transition into or out of employment in any one month. However, when considered over a longer period, the percentages accumulate, and where gaps between lone and couple mothers exist, they will become more apparent over time. For example, when calculated over this sample, for lone or couples mothers who were not employed in one month, 3.6% were employed the following month. If a 12-month period is used instead of the one-month period, 26% of those not employed at the beginning of this period had entered employment at some time in the following 12 months. For employed mothers, there were different monthly rates of leaving employment for lone and couple mothers (2.0% and 1.3% respectively). Over a 12-month period, this translated into a gap of 5% between lone and couple mothers (15% of employed lone mothers and 10% of employed couple mothers left employment over 12 months).

To use these findings about transition rates to explain why lone and couple mothers' employment rates are different, it is necessary also to take account of the proportions of lone and couple mothers "at risk" of making a transition into or out of employment. Table 4 showed that, on average, 58% of lone mothers and 68% of couple mothers were employed in one month. This greater "stock" of couple mothers who are employed, compared to lone mothers who are employed, is sustained over time, given the higher exit rates from employment for lone mothers. Conversely the "stock" of not-employed mothers is greater in any one month as a percentage of lone mothers than couple mothers (Table 4). Because the transition rates into employment are low, and are the same for lone and couple mothers, again, the lower rates of employment of lone mothers are perpetuated. Of course, this is a simplification of the transitions that occur, since mothers can transition into and out of lone parenthood as well as into and out of employment.

We could ask when or how this disparity in employment rates between lone and couple mothers arises. Does it occur after women become single mothers, either through separation or becoming parents while single? Or do women who become lone mothers already have lower employment attachment even before they become single parents? An indication that lone mothers have relatively low employment rates even before they become mothers can be revealed from the Longitudinal Study of Australian Children. Using these data, Baxter and Gray (2008) reported that 79% of mothers with only one child had been employed during the pregnancy for that child. If these data are disaggregated by mothers' relationship status, we find that of these mothers with one child, 65% of lone mothers and 81% of partnered mothers were employed during their pregnancy with that child. That is, even before lone mothers become parents, their attachment to the labour market is weaker. Whether employment rates are lower prior to relationship separation for lone mothers is an empirical question that could be explored further with these LSAC or HILDA data.

The weaker connection to the labour market for lone mothers has been apparent throughout this report, for both employed and not employed mothers. This was evident in the different characteristics of lone and couple mothers (Table 3), as well as in the detailed analyses of labour market characteristics of employed and not employed lone and couple mothers.

Exploration of the impact of characteristics other than relationship status on mothers' employment transitions revealed that several factors, in particular education levels, previous work experience and health status, were important in explaining differences in transitions into and out of employment These findings were especially important with regard to lone mothers' employment, given that, on average, lone mothers had lower levels of education, less work experience and poorer self-reported health than couple mothers. Such differences may contribute to the finding that lone mothers were not more likely than couple mothers to enter employment, despite their being more likely to be unemployed (indicating active job search), more likely to want a job, and more likely to be available to start work.

While our analyses included a range of characteristics of mothers in order to isolate differences in employment transitions related to relationship status, it is possible that lone and couple parents differed on other characteristics not included in these analyses. Butterworth (2003), for example, showed that for income support recipients in Australia, lone mothers had a greater incidence of mental health problems, substance abuse disorders and a history of having experienced physical or sexual violence than couple mothers. These factors may all contribute to lone mothers' difficulties in engaging with the labour market. Our finding of a higher exit rate from employment for lone mothers may therefore not just relate to lone parenthood, but to such differences in the characteristics of lone and couple mothers.

Related to this, the higher transition rate from employment for lone mothers may reflect differences in the types of jobs undertaken by lone and couple mothers. People leaving income support for employment, or combining income support with employment, often work in precarious jobs with low wages and with limited employment benefits, such that their short-term financial gains from employment are not always evident (Bodsworth, 2009; Walters, 2002b). There is also the question of whether these jobs are a stepping stone to better jobs (such that "any job is a good job"), or whether they are part of a "low pay/no pay cycle" (Evans et al., 2004; Fok et al., 2009). The analyses in this report showed lone mothers were more likely than couple mothers to be employed in casual jobs, and to have a more tenuous connection to the labour market. It is possible that lone mothers' higher rate of transition out of employment is related to these labour market characteristics. More detailed analyses of the types of jobs undertaken by lone compared to couple mothers, their reasons for leaving employment and longer term employment trajectories would be useful to further our understanding of the results presented in this report.

Policy implications

As discussed earlier in this report, a focus of government policy in Australia, as in other industrialised countries, has been on providing incentives for lone mothers to participate in the labour market. An important policy implication of the research presented in this report is that a significant reason for the lower employment rate of lone mothers is their higher exit rate from employment, and thus there is a potential role for policy in improving employment retention for lone mothers. Given this finding, this section provides an overview of the range of policies that may potentially play a role in improving employment retention of lone mothers. Other OECD countries, especially the UK and the US, have begun to recognise job retention as an important policy focus (Bell et al., 2006; Evans et al., 2004; Kellard et al., 2002). Hirsch (2006), writing about possible approaches in the UK to improving lone parents' in-work outcomes, including job retention but also job advancement, suggests three broad approaches: "sustainable entry strategies", "support for new entrants" and "improving jobs".

"Sustainable entry strategies" are concerned with enabling and encouraging take-up of sustainable jobs through appropriate training and job search assistance. This addresses the fact that many lone parents, on entering employment, often take up precarious jobs; for example, short-term, low-paid, temporary and casual jobs. If, instead, lone mothers are encouraged or helped to take up jobs with better employment conditions or a better fit to their caring responsibilities, this might result in greater employment retention and therefore less churning back into non-employment. This is not without contention, however, as such jobs may be difficult for lone mothers to attain given their relatively low levels of human capital, and this then relates back to the question of whether any job is better than no job. Is it better for lone mothers to move into and out of different, relatively low status jobs, if the alternative is to remain longer out of employment while waiting for a more sustainable job?

"Support for new entrants" covers the provision of continued case management and support to lone parents after commencement of employment, as well as access to training; for example, the UK New Deal for Lone Parents, which includes access to advice and support in the period following entry into work (Millar & Ridge, 2009). Further work is underway in the UK and the US to explore policies that might result in better job retention, notably through their Employment Retention and Advancement programs. While not yet fully evaluated, in the UK, these programs included job coaching and mentoring; a retention bonus for staying in work for a specified minimum number of hours and weeks; training fees and bonuses; and services to assist parents to increase their earnings or to find a higher paying job (Bell et al., 2006; Kellard et al., 2002).

Support for new entrants can also include wage supplements or tax credits to help address the fact that some employed lone parents may earn quite low incomes, and may actually be no better off financially than if they were to remain out of employment, because of the withdrawal of benefits once income reaches a certain threshold. As a result, "making work pay" is a very significant focus of policy in relation to lone parents' employment across the OECD countries (Knijn, Martin, & Millar, 2007; Millar, 2008).

"Improving jobs", in Hirsch's (2006) framework, is related to wages, employment conditions and job protection for workers, especially those in more vulnerable jobs. In addition to access to basic employment conditions (such as reasonable wages), access to family-friendly working conditions is likely to be especially relevant to mothers of young children, which, for example, has been addressed in the UK in relation to parents' rights to request flexible work hours (Millar & Ridge, 2009). Demand-driven policies to improve job retention are particularly relevant here, including those in which effort is made to identify the demand for jobs in the local labour market through consultation with employers, in order to provide training opportunities and match jobseekers to those jobs (Kellard et al., 2002).

These policy approaches would extend the current focus on moving lone mothers into employment. Like Australia, most countries have policies that encourage transitions into employment, through "activation strategies", or policies that require mothers, once children reach a certain age, to participate in work, in job search, or in other activities such as training or community work, in order to develop work-related skills. Also, lone and couple mothers' employment decisions and outcomes are affected by a broader range of policies, including those in the areas of child care and industrial relations. See, for example, Adema and Whiteford (2007), Evans et al. (2004), Knijn et al. (2007), Millar (2008), Millar and Ridge (2009), and Yeo (2007) for discussion of these policies.

Conclusion

In this paper we have explored the employment transitions of lone and couple mothers to determine whether differences in exits from employment or entrances to employment could explain the lower employment rates of lone mothers. We found that lone mothers had a somewhat higher rate of exiting employment in any one month, compared to couple mothers, but found no difference between lone and couple mothers in their rate of entering employment. The different rates of exiting employment remained significant when several characteristics of lone and couple mothers were taken into account, including educational attainment and extent of recent work history. Policy measures aimed at increasing the job retention of lone mothers may therefore help to increase the employment rate of lone mothers. However, the vast majority of not-employed lone mothers are remaining out of work rather than transitioning between non-employment and employment. Some of these mothers are likely to be focusing on providing full-time care to their children at this life cycle stage. Other lone mothers appear to want to work, but face substantial barriers to doing so. Providing additional tailored supports to these mothers may be another way of reducing the employment gap between lone and couple mothers.

References

• Acs, G., Loprest, P., & Roberts, T. (2001), Final synthesis report of findings from ASPE "Leavers" Grants . Washington: The Urban Institute.
• Adema, W., & Whiteford, P. (2007). Babies and bosses: Reconciling work and family life. A synthesis of findings for OECD countries. Paris: OECD.
• Alexander, M., Baxter, J., Hughes, J., & Renda, J. (2005). Evaluation of the impact of activity requirements for parenting payment customers on their children aged 13–15 years: The final report to the Department of Employment and Workplace Relations. Melbourne: Australian Institute of Family Studies.
• Australian Bureau of Statistics. (2003). Marriages and divorces Australia, 2002 (Cat. No. 3310.0). Canberrra: ABS.
• Australian Bureau of Statistics. (2009). Childhood education and care Australia, 2008 (Cat. No. 4402.0). Canberra: ABS.
• Australian Bureau of Statistics. (2010). Labour force, Australia (Cat. No. 6202.0). Canberrra: ABS.
• Australian Institute of Health and Welfare. (2008). Making progress: The health, development and wellbeing of Australia's children and young people. Canberra: AIHW.
• Avdeyeva, O. (2006). In support of mothers' employment: Limits to policy convergence in the EU? International Journal of Social Welfare, 15(1), 37–49.
• Barrett, G. (2002). The dynamics of participation in the sole parent pension. Economic Record, 78, 1–17.
• Baxter, J. (2008). Timing of mothers' return to work after childbearing: Variations by job characteristics and leave use (Research Paper No. 42). Melbourne: Australian Institute of Family Studies.
• Baxter, J., & Alexander, M. (2008). Mother's work–family strain in single and couple parent families: The role of job characteristics and supports. Australian Journal of Social Issues, 43(2), 195–214.
• Baxter, J., & Gray, M. (2008). Submission from the Australian Institute of Family Studies to the Productivity Commission Inquiry into Paid Maternity, Paternity and Parental Leave. Melbourne: Australian Institute of Family Studies.
• Baxter, J., Gray, M., Alexander, M., Strazdins, L., & Bittman, M. (2007). Mothers and fathers with young children: Paid employment, caring and wellbeing (Social Policy Research Paper No. 30). Canberra: Department of Families, Community Services and Indigenous Affairs.
• Bell, K., Branosky, N., Fizgerald, J., Greenberg, M., Harkness, S., Hirsch, D. et al. (2006). Staying on, stepping up: How can employment retention and advancement policies be made to work for lone parents? London: One Parent Families.
• Blau, F. D., & Ferber, M. A. (1992). The division of labor within the household. In F. D. Blau, & M. A. Ferber (Eds.). The economics of women, men, and work (2nd Ed.). Englewood Cliffs, NJ: Prentice Hall.
• Bodsworth, E. (2009). Making work pay and making income support work. Fitzroy, Vic.: Brotherhood of St Laurence, Fitzroy, Melbourne.
• Bradbury, B. (2003). Child poverty: A review (FaCS Policy Research Paper No. 20). Sydney: Department of Family and Community Services.
• Breunig, R., & Gong, X. (2010). Child care availability, quality and affordability: Are local problems related to maternal labour supply? (Treasury Working Paper 2010–02). Canberra: Department of The Treasury.
• Buddelmeyer, H., Wooden, M., & Ghantous, S. (2006). Transitions from casual employment in Australia. Melbourne: Melbourne Institute of Applied Economic and Social Research.
• Butterworth, P. (2003). Multiple and severe disadvantage among lone mothers receiving income support. Family Matters, 64, 22–29.
• Callegaro, M. (2007). Seam effects changes due to modifications in question wording and data collection strategies: A comparison of conventional questionnaire and event history calendar seam effects in the PSID. Lincoln: University of Nebraska.
• Cancian, M., Haveman, R., Kaplan, T., & Wolfe, B. (1999). Post-exit earnings and benefit receipt among those who left AFDC in Wisconsin (Special Report No. 75). Madison: Institute for Research on Poverty.
• Cass, B. (2006). Care giving and employment: Policy recognition of care and pathways to labour force return. Australian Bulletin of Labour, 32(3), 240–256.
• Centrelink (2003). A guide to Australian Government payments, 20 March–1 July 2003. Canberra: Centrelink.
• Centrelink. (2005). Centrelink information: A guide to payments and services, 2004–05. Canberra: Centrelink.
• Chalmers, J. (1999). Sole Parent Exit Study: Final report. Report to the Department of Family and Community Services. Canberra: Department of Family and Community Services.
• Dawkins, P., Gregg, P., & Scutella, R. (2002). The growth of jobless households in Australia. Australian Economic Review, 35(2), 133–154.
• Dex, S., Joshi, H., Macran, S., & McCulloch, A. (1998). Women's employment transitions around child bearing. Oxford Bulletin of Economics and Statistics, 60(1), 79–98.
• Eardley, T. (2001). Sole parents and welfare dependency. Just Policy: A Journal of Australian Social Policy, 21, 2–7.
• Evans, M., Eyre, J., Millar, J., & Sarre, S. (2003). New deal for lone parents: Second synthesis report of the national evaluation. London: Department of Work and Pension.
• Evans, M., Harkness, S., & Arigoni Ortiz, R. (2004). Lone parents cycling between work and benefits (Research Report 217). London: Department of Work and Pensions.
• Fok, Y., Jeon, S., & Wilkins, R. (2009). Does part-time work help or hinder lone mothers' movements into full-time employment? (Melbourne Institute Working Paper No. 25/09). Melbourne: Melbourne Institute of Applied Economic and Social Research, University of Melbourne.
• Funder, K. (1993). Women's post-separation employment and reliance on social security. In K. Funder, M. Harrison, & R. Weston (Eds.), Settling down: Pathways of parents after divorce (pp. 83–115). Melbourne: Australian Institute of Family Studies.
• Gong, X., Breunig, R., & King, A. (2010). How responsive is female labour supply to child care costs (Treasury Working Paper 2010–03). Canberra: Department of The Treasury.
• Gornick, J., & Meyers, M. (2003). Families that work: Policies for reconciling parenthood and employment. New York: Russell Sage Foundation.
• Gray, M., Qu, L., Renda, J., & de Vaus, D. (2006). Changes in the labour force status of lone and couple Australian mothers, 1983–2005. Australian Journal of Labour Economics, 9(4), 395–416
• Gregory, R. (2002, June). Can this be the Promised Land? Work and welfare for the modern woman. Paper presented at the National Institutes Public Lecture, Parliament House.
• Gregory, R., Klug, E., & Thapa, P. (2008). Cut-offs, knock-ons and welfare payment taper changes: An evaluation of the July 2000 tax and welfare changes for lone mothers. Economic Record, 84(266), 291–309.
• Harding, A., Vu, Q., Percival, R., & Beer, G. (2005). Welfare-to-work reforms: impact on sole parents. Agenda, 12(3), 195–210.
• Haynes, M., Western, M., Yu, L., & Spellak, M. (2008, 1–4 August). Employment transitions of Australian women: Analysing nominal data from a panel survey. Paper presented at the 103rd Annual meeting of the American Sociological Association, Boston, USA.
• Headey, B., & Verick, S. (2006). Jobless households: Longitudinal analysis of the persistence and determinants of joblessness using HILDA data for 2001–03 (Melbourne Institute Report No. 7) Melbourne: Melbourne Institute of Applied Economic and Social Research, University of Melbourne.
• Hirsch, D. (2006). Welfare in work: The missing link in welfare reform. In K. Bell, N. Branosky, J. Fitzgerald, M. Greenberg, S. Harkness, D. Hirsch et al. (Eds.), Staying on, stepping up: How can employment retention and advancement policies be made to work for lone parents? (pp. 11–16). London: One Parent Families.
• Hynes, K., & Clarkberg, M. (2005). Women's employment patterns during early parenthood: A group-based trajectory analysis. Journal of Marriage & Family, 67(1), 222–239.
• Jäckle, A. (2008). The causes of seam effects in panel surveys (Institute for Social and Economic Research Working Paper No. 2008–14). Colchester: University of Essex.
• Jäckle, A., & Lynn, P. (2004). Dependent interviewing and seam effects in work history data (Institute for Social and Economic Research Working Paper No. 2004–24). Colchester: University of Essex.
• Jeon, S. H. (2008). The impact of life cycle events on women's labour force transitions: A panel analysis. The Economic Record, 84(Special Issue), 83–98.
• Joesch, J. M. (1997). Paid leave and the timing of women's employment before and after birth. Journal of Marriage and the Family, 59(4), 1008–1021.
• Johnson, W., & Skinner, J. (1986). Labour supply and marital separation. The American Economic Review, 76(3), 455–468.
• Kellard, K., Adelman, L., Cebulla, A., & Heaver, C. (2002). From job seekers to job keepers: Job retention, advancement and the role of in-work support programmes. Leeds: Department of Work and Pensions
• Knights, S., Harris, M., & Loundes, J. (2000). Dynamic relationships in the Australian labour market: Heterogeneity and state dependence (Melbourne Institute Working Paper No. 6/00). Melbourne: Melbourne Institute of Applied Economic and Social Research.
• Knijn, T., Martin, C., & Millar, J. (2007). Activation as a common framework for social policies towards lone parents. Social Policy & Administration, 41(6), 638–652.
• McHugh, M., & Millar, J. (1996). Sole mothers in Australia: Supporting mothers to seek work (Discussion Paper 71). Sydney: Social Policy Research Centre.
• Millar, J. (2008). Making work pay, making tax credits work: An assessment with specific reference to lone parent employment. International Social Security Review, 61(2), 21–38.
• Millar, J., & Evans, M. (Eds.). (2003). Lone parents and employment: International comparisons of what works. Sheffield: UK Department for Work and Pensions.
• Millar, J., & Ridge, T. (2009). Relationships of care: Working lone mothers, their children and employment sustainability. Journal of Social Policy, 38(1), 103–121.
• Neilson, R., & Gottschalk, A. (2009). Estimating labour force transitions in the presence of a seam effect. Applied Economics Bulletin Peer-Reviewed Working Paper Series, 2, 1–19.
• Ondrich, J., Spiess, C. K., Yang, Q., & Wagner, G. G. (2003). The liberalization of maternity leave policy and the return to work after childbirth in Germany. Review of Economics of the Household, 1(1), 77–110.
• Paull, G. (2007). Partnership transitions and mothers' employment (Research Report No. 452). Leeds: Department of Work and Pensions.
• Perkins, D., & Scutella, R. (2008). Improving employment retention and advancement of low-paid workers. Australian Journal of Labour Economics, 11(1), 97–114.
• Richardson, S., & Miller-Lewis, L. (2002). Low wage jobs and pathways to better outcomes (Working Paper No. 02/29). Wellington: New Zealand Treasury.
• Rips, L., Conrad, F., & Fricker, S. (2003). Straightening the seam effect in panel surveys. Public Opinion Quarterly, 67, 522–554.
• Shaver, S., King, A., McHugh, M., & Payne, T. (1994). At the end of eligibility: Female sole parents whose youngest child turns 16 (SPRC Report No. 117). Sydney: University of New South Wales.
• Stromback, T., Dockery, A. M., & Ying, W. (1998). Transition in the labour market: Evidence from the Survey of Employment and Unemployment Patterns (Working Paper No. 1). Melbourne: Melbourne Institute of Applied Economic and Social Research.
• Toohey, M., & Beer, G. (2003, 9 July) Is it worth working now? Financial incentives for working mothers under Australia's new tax system. Paper presented to the 2003 Australian Social Policy Conference, Sydney.
• Walters, M. (2002a). Labour market participation and the married to sole mother transition. In T. Eardley, & B. Bradbury (Eds.), Competing visions: Refereed proceedings of the National Social Policy Conference 2001 (SPRC Report 1/02) (pp. 411–21). Sydney: Social Policy Research Centre, University of New South Wales.
• Walters, M. (2002b). Working their way out of poverty? Sole motherhood, work, welfare and material well-being. Journal of Sociology, 38(4), 361–380.
• Watson, N. (2009, July 16–17). Disentangling overlapping seams: The experience of the HILDA survey. Paper presented at the 2009 HILDA Research Conference, Melbourne.
• Watson, N., & Wooden, M. (2002a). The Household Income and Labour Dynamics in Australia (HILDA) survey: Wave 1 survey methodology (HILDA Project Technical Paper Series No. 1/02). Melbourne: Melbourne Institute of Applied Economic and Social Research, University of Melbourne.
• Watson, N., & Wooden, M. (2002b). Assessing the quality of the HILDA survey Wave 1 data (HILDA Project Technical Paper Series No. 4/02). Melbourne: Melbourne Institute of Applied Economic and Social Research, University of Melbourne.
• Whiteford, P. (2001). Lone parents and employment in Australia. In J. Millar, & K. Rowlingson (Eds.), Lone parents, employment and social policy: Cross-national comparisons (pp. 61–86). Bristol: The Policy Press.
• Wilson, K., Pech, J., & Bates, K. (1999). Parents, the labour force and social security (Policy Research Paper No. 2). Canberra: Department of Family and Community Services.
• Yeo, A. (2007). Experience of work and job retention among lone parents: An evidence review (Working Paper 37). London: Department for Work and Pensions.

Lists of tables and figures

List of tables

• Table 1 Summary of employment data for mothers with children aged under 15 years, by relationship status, Wave 2
• Table 2 Sample numbers and person-month records for mothers with children aged under 15 years, by year and relationship status, 2002-08
• Table 3 Characteristics of mothers with children aged under 15 years, by relationship and employment status, Waves 2-8
• Table 4 Month-to-month employment transitions of mothers with children aged under 15 years, by relationship status, pooled data, 2002-08
• Table 5 Month-to-month employment transition matrix of mothers with children aged under 15 years, pooled data, 2002-08
• Table 6 Multivariate analyses of the likelihood of leaving employment for mothers employed in the previous month, pooled monthly data
• Table 7 Multivariate analyses of the likelihood of entering employment for mothers not employed in the previous month, pooled monthly data

List of figures

• Figure 1 Employment rates in selected OECD countries of mothers aged 15-64 years, by relationship status, 2007
• Figure 2 Employment rates of mothers with children aged under 15 years, by relationship status, June 1984 to June 2008
• Figure 3 Employment rates of mothers with children aged under 15 years, by relationship status and age of youngest child, June 2000 to June 2008 (ABS)
• Figure 4 Employment rates of mothers with children aged under 15 years, by relationship status and age of youngest child, June 1984 to June 2008 (HILDA)
• Figure 5 Percentage leaving employment, by relationship status, mothers employed in the previous month, 2002-08
• Figure 6 Percentage leaving employment, by relationship status and recent work history, mothers employed in the previous month
• Figure 7 Percentage leaving employment, by relationship status and highest level of education, mothers employed in the previous month
• Figure 8 Percentage leaving employment by relationship status and age of youngest child, mothers employed in the previous month
• Figure 9 Percentage leaving employment by relationship status and number of children, mothers employed in the previous month
• Figure 10 Percentage leaving employment by relationship status and self-reported health status, mothers employed in the previous month
• Figure 11 Percentage leaving employment by relationship status and country of birth, mothers employed in the previous month
• Figure 12 Percentage entering employment by relationship status, mothers not employed in the previous month, 2002-2008
• Figure 13 Percentage entering employment by relationship status and recent work history, mothers not employed in the previous month
• Figure 14 Percentage entering employment by relationship status and highest level of education, mothers not employed in the previous month
• Figure 15 Percentage entering employment by relationship status and age of youngest child, mothers not employed in the previous month
• Figure 16 Percentage entering employment by relationship status and number of children, mothers not employed in the previous month
• Figure 17 Percentage entering employment by relationship status and self-reported health status, mothers not employed in the previous month
• Figure 18 Country of birth of those entering employment, mothers not employed in the previous month, by relationship status

Appendix A: Data creation and derivation

Seam effects

HILDA respondents are surveyed every 12 months and at each wave calendar data is collected for the 14-18 month period prior to the interview. This creates an "overlapping seam", or a period of time in which data were available from two waves. This overlap period was included to allow analysis of and reduction in the magnitude of seam effects, a common problem in panel data where transitions are over-reported between survey waves (see Callegaro, 2007; Jäckle, 2008; Jäckle & Lynn, 2004; Neilson & Gottschalk, 2009; Rips, Conrad, & Fricker, 2003; Watson, 2009).

The magnitude of errors across the seam varies between panel surveys and is largely dependent on the characteristics of the sample, the data collection methodology used, the variables being examined and the length of the reference period (Callegaro, 2007). In examining the magnitude of seam effects in the HILDA study, Watson (2009) found that, across the first seven waves, 17% of the job spells identified in the first report of the overlapping period do not match those in the second report, with 12% of the job spells dropped (that is, they were included in the first report and not in the second) and 5% misplaced (that is, the job spell was recorded as starting/ending at a different point in each report). In addition, 8% of the job spells reported in the second period were not reported in the first.

In these analyses, the job spell information reported closest to the period being recalled (the first report) was used, unless there was missing data (then the less recent report was used). To test the impact of our method of addressing seam effects on the results, the analyses were replicated using the less recently reported job spell information. The findings were consistent with those found using the most recent reports of job spell information.

Monthly relationship status

To derive relationship status at each month, indicators of relationship status at each wave were used, as well as responses to questions about when live-in relationships and marriages began or ended.

The structure of these questions on relationship changes varied slightly across waves. Unfortunately. the questions in Wave 1 placed limits on the number of relationship changes that could be reported. Also, for many respondents it was unclear what their relationship status would have been at 1 July 2000, the earliest date in the calendar data. Therefore, there was no starting point from which to map relationship changes. This incomplete picture of relationships prior to the date of the Wave 1 survey contributed to the decision to exclude Wave 1 data from the analyses.

In the subsequent waves, respondents were asked about changes in marital status and live-in relationships that had occurred since their previous interview date, and the month in which these changes occurred. This information, along with relationship status at each interview, was used to develop indicators of respondents' relationship status at each of the calendar time periods.

A limitation of the method by which the relationship status information was collected in the HILDA survey is that there were restrictions to the number of relationships for which start and end dates were provided. The questions were most restrictive for Waves 1 and 2, but other waves also potentially had incomplete relationship data.

In Wave 2, respondents were asked if their marital status had changed and when the change happened. There was not scope for multiple marital transitions to be recorded (e.g., marrying then divorcing within one wave). In cases where two transitions occurred, the second transition would be picked up at the subsequent interview, therefore the relationship change would be identified later than when it happened, but not missed. For more than two marital status transitions, relationship changes will have been missed, although this would be very rare.⁸

In Waves 3 to 8, respondents were given the opportunity to provide separate dates for different types of marital status changes, including divorce, separation, being widowed and getting married. However, they only had the opportunity to report one of each type of change, therefore some relationship changes may be missed; however, the number missed would be very low. Therefore, overall, missed relationship changes are not likely to have any visible impact on the results, as the number of respondents affected is likely to have been very small.

Within-month employment

Collapsing the data into monthly periods results in a slight underestimation being made of the number of transitions into and out of employment.

Appendix Table A1 shows that both lone and couple mothers made very few within-month employment transitions and were instead usually employed for the whole month or not employed for the whole month. Within-month employment transitions took place in 1.6% of months for lone mothers and 1.3% of months for couple mothers. Depending on when these transitions took place, they may have been missed in the analysis when the data was converted to monthly periods.

Further examination reveals that 163 transitions out of employment would not have been picked up in the current analysis of monthly transitions; 63 for lone mothers and 100 for couple mothers. These include instances where respondents were employed in two consecutive months but not for the whole time or where they were employed at the start of a month, left in the middle, then were employed again at the end of the month. Considering that we are examining almost 150,000 person-months, the number of missed transitions is very small and not likely to have an impact on the results.

Note: Percentages may not total 100% due to rounding.

Appendix B: Supplementary tables

Note: Random effects models also included controls for missing health information and missing employment experience information. Post-estimation tests showed that none of the lone mother coefficients were significantly different from the couple mother coefficients. a Rho measures proportion of total variance contributed by person-level variance. *** p < .001, ** p < .05, * p < .01.

Note: Model also included controls for missing health information and missing employment experience information. Post-estimation tests showed that none of the lone mother coefficients were significantly different from the couple mother coefficients. a Rho measures proportion of total variance contributed by person-level variance. *** p < .001, ** p < .05, * p < .01.