Australian mothers' participation in employment

Analyses of social, demographic and family characteristics using the Household, Income and Labour Dynamics in Australia (HILDA) survey
Research Paper No. 52 – November 2013

3. HILDA data

HILDA is a nationally representative annual panel survey that commenced with Wave 1 in 2001 (Watson & Wooden, 2002). 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 across Australia. Members of 7,682 households completed interviews, resulting in 13,969 completed individual interviews and a response rate of 66%. While the number of participants from the original sample has declined over the waves due to attrition, at each wave new members to households are added in. Further, in Wave 11, the sample was topped up with an additional 2,153 households and 5,477 individuals. The purpose of this top-up sample was to address the fact that recent arrivals to Australia were no longer well represented in the HILDA sample (Watson, 2012). The top-up sample came from the general population, and so boosted sample sizes for Australia-born as well as immigrant respondents.

At each wave, to Wave 10, the sample has included approximately 2,000 mothers of children aged under 15 years, with somewhat more in Wave 11, due to the top-up of the sample.7

The data from Wave 1 (2001) to Wave 11 (2011) were combined to compare maternal employment rates from each of the waves of the study. Use of HILDA in this way (treated as repeated cross-sectional analyses) is not the preferred approach to analyse trends, as changes in the composition in the sample across waves are not taken into account (except through the application of sample weights). This may be an important factor particularly with Wave 11, with the addition of the top-up sample. Watson (2012) explored whether this made a difference to estimates produced using HILDA, and found that the inclusion of the top-up sample brought most of the estimates compared closer to those produced from Australian Bureau of Statistics (ABS) surveys. It is important, though, to be mindful that estimates of the subsample of HILDA comprising mothers with children aged under 15 years may be affected by the changing composition of the sample through wave-to-wave attrition and the top-up of the sample.

This research focuses on mothers with children aged under 15 years, who are referred to simply as "mothers" throughout the report. Information about relationships between household members is used to classify these women as being lone or couple mothers at the time of each interview, with lone mothers being those who do not have a co-resident partner at the time of the interview. That is not to say these mothers have always been lone or couple mothers, and in particular, many of the lone mothers will have previously been partnered. At Wave 11, there were 502 lone mothers and 2,067 couple mothers with children aged under 15 years.

Employment status here is derived using standard labour force definitions, such that a mother is counted as being employed if she undertakes at least one hour of paid work in the week before the survey (see ABS, 2007). Mothers are also counted as being employed if they are on leave from a job, but have been away for fewer than 4 weeks, or have been away for longer than this but in the last 4 weeks have received pay at some stage (i.e., they are on paid leave).

Mothers on longer term unpaid leave are counted as being either unemployed or not in the labour force (NILF), depending on their answers to questions about job search and their availability to start work. Unemployed mothers are those who are actively seeking work and available to start work. Mothers are classified as being not in the labour force if: (a) they do not want to work; (b) they want to work, but are not seeking work, regardless of whether they are available to start work; and (c) they are seeking work but are not available to start work. Responses to these labour force questions are examined in this report to explore differences in levels of attachment to the labour force and potential barriers to employment.

In addition to mothers' employment status at the survey date, this paper also makes use of data collected in an "employment calendar" for each wave of HILDA. For these data, respondents are asked about their work and study activities for the period starting from 1 July of the previous year, up to the survey date.8 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. Derived information, based on these data, describes participation in employment in the financial year prior to the survey date.9 This includes information on what percentage of that year was spent in employment.

Mothers were assigned to one of three categories according to the percentage of time they spent working: (a) employed for little/none of the year (0-9% of the year); (b) employed for part of the year (10-89%); or (c) employed for most/all of the year (90-100%). Mothers employed for part of the year include those with intermittent or casual employment, but also includes those who left or started employment part way through the year. This latter group with part-year employment was quite small and could not be disaggregated further (see results in Table 1).

The analyses presented here begin with an overview of trends, with data sourced from Waves 1 to 11 of HILDA, using the Wave 11 HILDA release. Most other analyses use Wave 11 (2011) data.

In the final section, which explores employment transitions, characteristics of mothers were taken from the Wave 10 (2010) survey and analysed according to their employment status one year later, as identified in Wave 11. For this section, some data from earlier waves were used instead, when items of interest had not been collected in Wave 10. For those items, the most recent year in which they had been collected was used. Some were from 2008 (e.g., work-family attitudes), others from 2007 (e.g., measures of sense of personal autonomy).

The choice of characteristics to be examined in these analyses was guided by the literature on the determinants of maternal employment. This led to the inclusion of variables such as age of youngest child, education level and health status. Additional characteristics were examined to explore relationships between more subjective measures of wellbeing and maternal employment. Associations between work-family attitudes and maternal employment were also explored, given the possible role of preferences in explaining mothers' participation in employment.

Wave-specific cross-sectional sample weights were used to adjust for non-response in the calculations of means and percentages. Statistical tests of differences in means (t-tests) and distributions (chi-square tests) were applied to unweighted data. Statistically significant differences (at p < .05) have been indicated throughout the report.


7 In Wave 11, there were 1,936 mothers from the prior sample and 633 from the top-up sample.

8 Each survey wave begins in August and almost all surveys are completed before the end of December.

9 Financial year data do not necessarily capture very recent work history for those who were interviewed some months into the new financial year. It is possible to derive other calendar data pertaining to the twelve months prior to the interview, but for this report, due to the complexity involved in undertaking these derivations, attention is restricted to the previous financial year (for which derived measures are provided with the dataset).